Skip to content

Posts from the ‘Agentic AI’ Category

Open-weight AI models are four months behind and 9x cheaper

Chart of open-weight AI models and closed models plotting Epoch Capabilities Index score against blended OpenRouter price per million tokens for 11 models
Kimi K2.6 scores 151.6 on the Epoch Capabilities Index at a blended $1.71 per million tokens. GPT-5.5 scores 158.2 at $11.25. Sources: Epoch AI, Epoch Capabilities Index, May 29, 2026; OpenRouter list prices, Nov. 14, 2026; Gartner, Strategies to Compete With Free Open-Weight AI Models (G00864526); analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Kimi K2.6 is the strongest of the open-weight AI models on Epoch AI’s Capabilities Index. It scores 151.60 and lists on OpenRouter at $0.95 per million input tokens and $4.00 per million output tokens. GPT-5.5 scores 158.22 and lists at $5 and $30. A lead of 6.6 index points costs 6.6 times as much on a blended basis.

A Gartner note published Nov. 12 explains that spread. Epoch AI measures the best open-weight models trailing the closed frontier by an average of four months and eight index points since January 2026. Gartner puts the lag at 12 months at the start of 2025 and under four months today, and it assumes one month by 2029. With the free model only four months behind, a per-token markup for raw intelligence gets hard to defend.

Gartner’s sharper point is about who pays once the license fee disappears. It calls free open models an illusion, because the host absorbs the compute bill, the MLOps payroll and all legal and compliance liability. For security and risk leaders, the open-weight decision is therefore a liability decision, and engineering teams often make it without pricing it that way.

I built this analysis from Epoch AI’s open versus closed capability data (May 29, 2026), OpenRouter list prices for 11 models on Nov. 14, 2026, and Gartner’s Strategies to Compete With Free Open-Weight AI Models (G00864526, Nov. 12, 2026). Blended prices, ratios and medians are my calculations.

Key findings

  • The best open-weight AI models trail the frontier by four months and eight capability points. Epoch AI measures that average since January 2026, and calls eight points similar to the gap between GPT-5 and GPT-5.5.
  • Gartner expects the lag to fall to one month by 2029. Gartner puts it at 12 months at the start of 2025 and under four months today, so about two thirds of it closed in under two years.
  • Closed models cost nine times more at the median. Across 11 models priced on OpenRouter, the median closed model blends to $6.69 per million tokens and the median open-weight model to $0.74.
  • Frontier output tokens cost up to 200 times more. GPT-5.5 lists output at $30 per million tokens, against $4.00 for Kimi K2.6 and $0.15 for gpt-oss-120b.
  • Self-hosting moves all legal and compliance liability to the buyer. Gartner says OpenAI and Anthropic indemnify outputs and IP claims by contract, and neither yet guarantees agentic transactions.
  • Gartner expects 55% of AI platform providers to sell outcome-priced tiers by 2029. Every outcome guarantee needs telemetry to prove it, and that telemetry is a data flow security teams have to review.

Open-weight AI models sit four months and eight points behind

Scatter chart of Epoch Capabilities Index scores by release date for 16 LLMs released from May 2025 to May 2026, six closed and ten with downloadable weights
Six closed models score between 154 and 158 on the index. The ten open-weight models score between 141 and 152, led by Kimi K2.6. Source: Epoch AI, Epoch Capabilities Index, May 29, 2026; analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Epoch’s index rolls many benchmarks into one capability score. The chart plots 16 models from that index, six closed and ten open-weight, released between May 2025 and May 2026. All six closed models arrived between February and May 2026 and score between 154.28 and 158.22. GPT-5.5 leads at 158.22, followed by Gemini 3.5 Flash at 156.31 and Claude Opus 4.7 at 156.18. By comparison, the strongest open-weight model, Kimi K2.6, scores 151.60. Kimi K2.5 follows at 148.22 and MiniMax-M2.5 at 147.44.

Averaged across the 16, closed models score 156.1 and open-weight models 145.9. The 16 are my selection, however, and they understate the top of the closed field. For example, the sample leaves out Epoch’s highest score, GPT-5.5 Pro at 159.35, and five other closed models released in 2026 that outscore Kimi K2.6. Epoch’s own average gap since January 2026 is eight points, which it calls similar to the gap between GPT-5 and GPT-5.5.

Gartner says the July 2026 launch of Kimi K3 signaled that open models could go neck and neck with closed ones on long-context reasoning, complex agentic workflows and tool execution. Epoch’s May 29 analysis predates that launch, so the four-month figure describes the field before Kimi K3 arrived.

Gartner says open-weight models closed two thirds of the lag in under two years

Bar chart of the time open-weight AI models take to match closed models: 12 months in early 2025, four months in 2026 and Gartner’s assumption of one month by 2029
Gartner puts the lag at 12 months in early 2025 and under four months today, and assumes one month by 2029. Epoch AI’s average since January 2026 is four months. Sources: Gartner, Strategies to Compete With Free Open-Weight AI Models (G00864526); Epoch AI, May 29, 2026; analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Gartner puts the average time for open-weight models to match closed ones at 12 months at the start of 2025 and under four months today. Epoch’s average since January 2026 is four months. Gartner’s strategic planning assumption is one month by 2029, and that figure is an assumption rather than a measurement.

The two sources do not measure the lag the same way. Epoch notes that its October 2025 analysis found an average of three months for January 2023 through October 2025, well below Gartner’s 12-month starting point. On Epoch’s figures, in other words, the lag has held between three and four months. The two-thirds decline is therefore Gartner’s estimate, and Epoch’s data does not show it.

A one-month lag changes procurement math. At that pace, any capability a closed provider ships would reach open weights before most enterprises finish a pilot. Gartner’s read is that competing on raw model intelligence stops being a sustainable strategy for providers, and the value moves to orchestration, context and trust.

Frontier output tokens cost up to 200 times more than open-weight models

Bar chart of OpenRouter list prices per million output tokens for closed models and open-weight AI models, from $30 for GPT-5.5 to $0.15 for gpt-oss-120b
GPT-5.5 lists output at $30 per million tokens. Kimi K2.6 lists at $4.00 and gpt-oss-120b at $0.15. Source: OpenRouter list prices, Nov. 14, 2026; analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

OpenRouter list prices show how wide the spread runs. GPT-5.5 lists at $5 per million input tokens and $30 per million output tokens. Claude Opus 4.7 lists at $5 and $25, and Gemini 3.5 Flash at $1.50 and $9. On the open side, Kimi K2.6 lists at $0.95 and $4.00 and GLM-5 at $1.00 and $3.20. Kimi K2.5 lists at $0.49 and $2.50, GLM-4.7 at $0.40 and $1.75, MiniMax-M2.5 at $0.27 and $0.95, DeepSeek-V3.2 at $0.29 and $0.43, and gpt-oss-120b at $0.03 and $0.15.

I then blended each price at three input tokens per output token. The median closed model comes to $6.69 per million tokens and the median open-weight model to $0.74, a 9x gap. Grok 4.20 breaks the pattern, though. It lists at $1.25 and $2.50, a blended $1.56 that undercuts Kimi K2.6 at $1.71, so closed does not always mean expensive.

These are list prices, so they overstate what many buyers pay. Caching and provider discounts lower the bill, and prices vary by hosting provider for the same open model. On Nov. 14, for example, OpenRouter showed temporary discounts of 59% on Kimi K2.6, 40% on GLM-5 and 28% on DeepSeek-V3.2, and I used the undiscounted rates. None of them include the cost of running the model yourself, which is where Gartner’s argument starts.

Free weights move the bill to the buyer

Table showing who carries compute, MLOps talent, liability and indemnity under self-hosted weights, a managed API and a closed frontier API
Self-hosting moves compute, talent and all legal and compliance liability to the buyer, as Gartner describes it. Source: Gartner, Strategies to Compete With Free Open-Weight AI Models (G00864526); analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Gartner lists three costs that transfer to an enterprise hosting open weights. They are compute infrastructure to serve the model, MLOps talent to run it, and all legal and compliance liability for what it produces.

The third cost is the one security leaders should own. Gartner says OpenAI, through its service terms, and Anthropic, through expanded commercial API protections, provide contractual indemnification for generated outputs and IP claims. It also notes that neither provider yet financially guarantees agentic transactions, and treats today’s copyright indemnity as a sign of where risk absorption is heading.

Download the weights and the indemnity stays behind. If an agent built on a self-hosted model takes an action that triggers a claim, then the claim lands on the enterprise. Gartner adds that managed open-weight API providers such as OpenRouter offer far cheaper inference than frontier models and are adding premium compliance features, including sovereign AI and data localization.

Gartner’s advice to model providers follows directly. Stop selling model capability and sell enterprise risk absorption instead, bundling API use with explicit financial indemnification for agentic transaction errors, performance guarantees and compliance tooling. In effect, the premium buys the use of someone else’s balance sheet.

Routing becomes the control plane

Gartner expects standalone orchestration middleware to shrink as open models handle multistep reasoning and tool use natively. What survives is hybrid routing that weighs cost, latency and liability for each request. It names Portkey, Cloudflare AI Gateway and OpenRouter as platforms that send complex reasoning to a closed frontier model and routine background work to open models.

Palo Alto Networks agreed to acquire Portkey on April 30, 2026 and completed the deal on May 29, calling it a centralized control plane to manage and protect autonomous AI agents. The April announcement says Portkey already processes trillions of tokens a month and gives access to more than 3,000 LLMs and MCP tools. My read is that a security platform vendor bought the gateway because it sees every prompt, every model selection and every agent action in one place. I covered how the large platforms are pulling ahead in the vendor tape analysis.

For a CISO, that reframes the open-weight question. The choice is between model selection that happens inside a control plane you can audit and model selection buried in application code you cannot see.

Gartner’s strategy map runs across three layers

Table of Gartner’s eight strategies for competing with free open-weight AI models across model providers, orchestration platforms and SaaS providers
Eight strategic pivots Gartner recommends across the model, orchestration and application layers, condensed. Source: Gartner, Strategies to Compete With Free Open-Weight AI Models (G00864526); analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Gartner’s note maps eight pivots across three layers of the AI stack. Model providers get four of them, and three of the four depend on assets an open-weight model cannot download, whether a balance sheet to underwrite risk, custom silicon or live enterprise data. Orchestration platforms get one, hybrid routing. SaaS providers get three, each built on something the customer has and the model does not, such as outcomes, perimeter control or workflow telemetry.

Each pivot also carries a market dependency in Gartner’s telling. Risk absorption depends on whether nondeterministic failures can be priced actuarially. Customer-controlled deployments depend on buyers funding GPU clusters and MLOps staff. Telemetry moats last only as long as foundation models do not absorb the same features from below.

Outcome pricing needs a measurement you trust

Gartner’s second planning assumption is that by 2029, 55% of AI platform providers will offer outcome-priced tiers with a completion record as a product feature. Its live example is Cognition, which guarantees Devin will deliver at least as much engineering value as a customer consumes and issues usage credits of up to $10 million when it falls short. Cognition measures value in equivalent engineering hours, using an estimator that runs on every completed task.

Gartner also points to Harvey fine-tuning open models inside a client’s virtual private cloud for zero data retention, and to Cursor logging micro-interactions such as cursor position and tab-accept rates to build a telemetry moat. It is candid that defining success for nondeterministic output is hard, and that proving the value of something subjective is a different problem from counting engineering hours.

The security version of that problem is sharper. A vendor that guarantees an outcome has to instrument the work to prove it, and instrumentation means telemetry leaving the environment. Every outcome guarantee is also a data flow, and procurement should review it as one.

Sixteen open-weight and closed AI models on the index

Table of 16 LLMs ranked by Epoch Capabilities Index score, six closed and ten with downloadable weights, with release date and OpenRouter input, output and blended prices
Sixteen models scored by Epoch AI, with OpenRouter list prices for the 11 that are priced in this analysis. Sources: Epoch AI, Epoch Capabilities Index, May 29, 2026; OpenRouter list prices, Nov. 14, 2026; analysis by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Epoch publishes an uncertainty range with each score, which the table leaves out. Several models appear in Epoch’s data at more than one reasoning-effort setting, and the table uses the setting Epoch lists in its comparison. Five of the 16 are not priced here, so the price medians rest on 11 models.

What security leaders should do about open-weight AI models

  • Find every self-hosted open-weight model already running in your environment. Then find who signed off on absorbing its liability. Often nobody did, because nobody framed it as a liability decision.
  • Treat the AI gateway as a security control. Routing decisions about which model sees which data belong in a control plane with an audit trail, and Palo Alto Networks’ Portkey acquisition shows where the market is putting that control. Gartner’s AI Governance Hype Cycle groups AI gateways with AI agent identity under agentic AI oversight, which I covered in 34% of enterprises govern AI with policies they only partly follow. Gartner’s first AI Governance Hype Cycle shows CISOs what to fund first..
  • Read frontier contracts for what is actually indemnified. Gartner’s reading is that output and IP claims are covered today and agentic transaction errors are not. That line matters most once agents start taking actions.
  • Price open weights on total cost. A 9x list-price advantage at the median shrinks once compute, MLOps staff and uninsured liability go into the comparison.
  • Ask every outcome-priced vendor what it measures and where that data goes. The guarantee is only as safe as the telemetry behind it.

Frequently asked questions

How far behind closed AI models are open-weight models?

Epoch AI measures the best open-weight models trailing the closed frontier by an average of four months and eight points on its Capabilities Index since January 2026. Gartner expects the lag to fall to one month by 2029.

How much cheaper are open-weight models?

On OpenRouter list prices from Nov. 14, 2026, the median open-weight model in this sample blends to $0.74 per million tokens and the median closed model to $6.69, about nine times more.

Are open-weight AI models free?

The weights are free to download. Gartner says the enterprise hosting them pays for compute and MLOps talent and absorbs all legal and compliance liability.

Do AI providers indemnify customers?

Gartner says OpenAI and Anthropic provide contractual indemnification for generated outputs and IP claims, and that neither yet financially guarantees agentic transactions.

What is outcome-based AI pricing?

Pricing tied to delivered results rather than seats or tokens. Gartner expects 55% of AI platform providers to offer outcome-priced tiers by 2029, and cites Cognition’s guarantee of up to $10 million in credits for Devin.

How I built this analysis

Capability scores and release dates come from Epoch AI’s open versus closed data insight, published May 29, 2026 by Jack Edwards and Luke Emberson. Open or closed status follows Epoch’s release-access labels, except for Gemini 3.5 Flash, which carries no label in Epoch’s data and which I treat as closed. The 16 models are my selection of six closed and ten open-weight releases, so the sample averages are illustrative. The four-month and eight-point figures are Epoch’s own. The 12-month, under-four-month and one-month lag figures and the 55% outcome-pricing assumption are Gartner’s, from G00864526.

Prices are OpenRouter list prices per million tokens captured Nov. 14, 2026 from each model’s page, before caching and discounts. Where a page showed a temporary discount, as the Kimi K2.6, GLM-5 and DeepSeek-V3.2 pages did, I used the undiscounted list rate. Blended price weights input three to one over output. Medians, ratios and averages are my calculations, and the 11 priced models are a sample rather than the full market. Adding or removing a model moves the median ratio.

More analysis from this blog

This post is my personal analysis of Gartner’s research and public sources and does not represent my employer.

Sources

Only 16% govern third-party AI: Gartner’s nine cybersecurity trends for 2027

Enterprises will spend $276.1 billion on information security in 2027. Gartner’s nine cybersecurity trends for the year explain where that money is already failing to keep up. Six of the nine trace to one cause. AI agents, frontier models and the machine identities they borrow are entering production faster than the controls built to govern, authenticate, test and train for them.

The readiness numbers behind the trends are blunt. For example, only 16% of organizations enforce workflows governing employee use of third-party AI applications. Likewise, 34% manage AI agent identity with tools designed for humans, and 15% have no approach at all. Fifty-one percent have not begun any post-quantum cryptography work, and 79% report no change to their PQC budget. Even so, Gartner dates the end of safe asymmetric encryption to 2029 through 2030. In its Cybersecurity Executive Outlook, 4Q26 (G00861818), it also predicts at least 200% higher migration costs for organizations that have not piloted PQC by 2027.

The budget is moving before the governance does

The three fastest-growing lines in Gartner’s 3Q26 security forecast are cloud security posture management at 29.1%, cloud access security brokers at 25.0% and zero trust network access at 22.3%. All three fund exactly the trends with the lowest readiness scores, while the legacy network controls they replace shrink. This analysis maps the nine trends in Gartner’s Top Trends in Cybersecurity, 2H26 (G00860263) to the 2027 forecast line by line. The pattern is consistent: the money is chasing the gaps, but the people and controls have not caught up.

Butterfly chart of Gartner's nine 2027 cybersecurity trends: on the left, the share of organizations still lacking the control each trend calls for, from 84% to 49%; on the right, 2026 to 2027 growth of the forecast lines that fund the response against the 11.6% market rate, with identity and SOC lines below market
The share of organizations still lacking each control, against 2027 growth of the forecast lines that fund the fix. Identity, SOC and consulting lines trail the 11.6% market rate despite gaps of 50% or more. Sources: Gartner, Top Trends in Cybersecurity, 2H26 (G00860263); 2027 Planning Guide for Cybersecurity (G00856222); Cybersecurity Executive Outlook, 4Q26 (G00861818); CISO Agenda for 2027 (G00859134); Forecast: Information Security, Worldwide, 2024-2030, 3Q26 (G00862059). Chart by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Why the list grew from six trends to nine

In February I covered the six trends in Gartner’s 2026 cybersecurity trends list. Since then, it has stayed one of the most-read pieces on this blog all year. Gartner published its 2026 trends press release on February 5, 2026, and has not published a 2027 edition. So this is not a rewrite of a report that does not exist yet. Instead, it is built from the three documents CISOs are using to set 2027 budgets right now. The mid-year update to the trends list, Top Trends in Cybersecurity, 2H26 (G00860263, August 19, 2026), expands six trends to nine. The 2027 Planning Guide for Cybersecurity (G00856222, October 7, 2026) then narrows those into four planning areas. The Forecast: Information Security, Worldwide, 2024-2030, 3Q26 (G00862059, September 24, 2026) prices them. Where the three disagree on emphasis, I say so.

Horizontal bar chart of Gartner's nine cybersecurity trends for 2027 grouped in three themes and colored in three shades of blue by theme, ranked within each theme by the survey statistic behind it, with each bar tagged as a threat signal or a readiness gap
Gartner groups its 2H26 cybersecurity trends into three themes; each bar is the survey finding Gartner cites to explain the trend, ranked within its theme. Source: Gartner, Top Trends in Cybersecurity, 2H26 (G00860263). Chart, ranking and threat-versus-readiness classification by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

One trend from February is gone: global regulatory volatility dropped off the list. A second trend dropped with it. Collaborative data security governance appeared in the full 1H26 report but never made the six-trend press release. Three new trends replaced them. Software supply chain security for AI components, technical debt as a barrier to AI, and frontier AI models rewriting vulnerability management. The removals matter as much as the additions. Gartner folded regulation and data governance into resilience and the CISO’s remit rather than treating them as standalone forces.

The spending backdrop. $276.1 billion and the growth rate just peaked

Gartner’s 3Q26 forecast puts worldwide end-user spending on information security at $247.5 billion in 2026 and $276.1 billion in 2027. That is an 11.6% increase in current U.S. dollars by my calculation from the forecast dataset. By comparison, Gartner’s own constant-currency figure for 2027 is 11.2%. Either way, 2026 was the peak growth year at 13.6%. Growth then decelerates every year, landing at 9.9% in 2030 when the market reaches $372.8 billion. I covered the 3Q26 forecast in detail in October, including why post-quantum firewall refreshes are the only hardware upgrade Gartner expects through 2030.

Stacked bar chart of worldwide information security spending 2024 to 2030 by security software, security services and network security with totals and growth rates, plus a line chart of each market's share of total spending
Growth peaked in 2026 at 13.6% and decelerates every year to 9.9% in 2030. Security software adds $17.0 billion in 2027, more than security services and network security combined, and its share of total spending rises from 49.6% in 2024 to 55.6% in 2030 while services fall from 39.3% to 33.6%. Source: Gartner, Forecast: Information Security, Worldwide, 2024-2030, 3Q26 (G00862059). Totals, growth rates and shares computed by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Software takes share from services

Security software carries the market. It grows 13.4% to $143.9 billion in 2027 and adds $17.0 billion of the $28.7 billion in new spending. Its share of total security spending rises from 49.6% in 2024 to 52.1% in 2027 and 55.6% in 2030, while security services fall from 39.3% to 33.6%. In other words, enterprises are substituting software for outside help. Services still grow 8.7% to $101.7 billion, but at the slowest rate of the three markets.

Network security grows 13.2% to $30.6 billion. Gartner also raised its long-term network outlook from a 9.8% to a 10.9% CAGR through 2030 on the strength of PQC-driven firewall replacements. Cloud security remains the fastest subsegment at 25.1% growth in 2027, reaching $20.8 billion. The securing-AI market Gartner tracks inside other security software grows from $3 billion in 2026 to $16 billion by 2030. Meanwhile, AI-amplified security products grow from $49 billion to $204 billion over the same period. I broke down the AI-amplified forecast in August.

Theme 1. Secure new frontiers

Trend 1. CISOs expand software supply chain security to cover AI models and agents

New to the list. Fifty-nine percent of CISOs rank third-party and supply chain attacks among their top five threats for the next 12 months. The figure comes from the 2026 Gartner CISO Role Survey (n=297) cited in the 2H26 trends report. Pressure to ship AI systems fast pushes teams onto open-weight and frontier models and the third-party components that wrap them. As a result, the security posture of the model itself usually goes unchecked. Gartner’s position is that point-in-time assessments and static controls cannot secure assets that update, drift and pull unverified dependencies continuously.

The 2027 Planning Guide adds the number that shows how far behind most programs are. Specifically, only 22% of organizations scan repositories for third-party component compromises that include models. Gartner’s actions are specific. Build an AI bill of materials and binary composition analysis into the CI/CD pipeline. Then extend third-party cyber-risk assessments to AI-specific software and its components. Finally, write incident response playbooks for AI supply chain disruptions that include isolating a service or deploying a workaround. Application security spending grows 10.6% to $10.7 billion in 2027, while vulnerability assessment within it grows 12.6%.

Trend 2. IAM adapts to secure and enable AI agents

Ninety-four percent of organizations report rising machine identity volumes over the past 12 months, per Gartner’s 2025 Machine Identity Survey (n=299). The growth is driven chiefly by AI and machine learning adoption (52%) and autonomous agents (35%). Fifty-eight percent have had at least one incident in which AI-linked machine identities were compromised. Those incidents caused moderate to significant business disruption at 42% of affected organizations. They also caused financial loss at 36% and reputational damage at 32%.

Gartner’s Theme: Cybersecurity at the AI Frontier (G00862284, November 3, 2026) puts the governance gap in two numbers from its 2026 Surviving the Agentic AI Era Survey. Thirty-four percent of organizations manage AI agent identity with tools designed for humans. Another 15% have no defined approach at all. The Cybersecurity Executive Outlook, 4Q26 goes further and finds 50% have no defined approach to limit AI agent access. Gartner’s Top Strategic Technology Trends for 2027, released at Gartner IT Symposium/Xpo in October, named AI agent identity one of the ten trends. It also predicted that by 2028, 25% of enterprise security breaches will trace back to AI agent abuse. The causes it names are unmanaged machine privileges and legacy credential practices.

However, the spending data says the money has not caught up. Identity and access management grows 8.6% to $25.5 billion in 2027, below the market rate. User authentication, the largest human-centric IAM line, grows 2.7%. By contrast, privileged access management grows 11.7% and identity governance 10.8%. The IAM budget is shifting from authenticating people to governing what machines are allowed to do. It is shifting slowly, though.

Trend 3. The quantum decryption threat advances while organizations lag

Gartner now dates a cryptographically relevant quantum computer to 2029 or 2030. Among CISOs who have started PQC work, fewer than 25% have made meaningful progress. That covers both a cryptographic inventory (24%) and a quantum risk assessment (22%). Sixty-five percent also say their board is unaware of or does not raise quantum threats, per the 2027 Planning Guide. The Cybersecurity Executive Outlook, 4Q26 sets the deadline in cost terms. Pilot PQC migration by 2027 or pay at least 200% more for the full migration later.

Still, budgets are moving ahead of CISOs here. The 3Q26 forecast expects at least 20% of customers in government, financial services and defense to upgrade firewalls to PQC-capable platforms in 2027. That share rises to more than 30% in 2028. The reason is that many installed firewalls cannot handle the processing demands of ML-KEM through a firmware update. Firewall equipment grows 13.7% to $22.4 billion in 2027. Encryption software, meanwhile, grows 10.4% to $1.7 billion. Gartner’s recommendation has not changed since February. Inventory cryptographic deployments, rank data by sensitivity and lifespan, and build crypto-agility so algorithms can be swapped without re-architecting systems.

Theme 2. Transform governance

Trend 4. Technical debt undermines organizational AI ambitions

New to the list. Forty-nine percent of cybersecurity leaders cite technical debt as a major barrier to executing their top functional priorities. Executive pressure to deploy AI forces teams onto unstructured data repositories without AI-ready controls. Oversharing and unauthorized access follow as a result. At the same time, frontier models are accelerating vulnerability discovery across infrastructure and enterprise software. Gartner names Anthropic’s Mythos, OpenAI’s Daybreak and Microsoft’s M-DASH. Each wave of discoveries then lands on legacy systems that cannot absorb the patches.

Gartner’s advice is to separate cybersecurity debt from enterprise technical debt. CISOs who try to own all of it will overwhelm their teams and open control gaps. The 2H26 report tells CISOs to inventory in-flight and short-term initiatives immediately. It then tells them to rank each one by how much risk it cuts from the dual threats of AI adoption and AI-discovered vulnerabilities. The Theme document is blunter. Closing the AI-Cyber Capability Gap will cost more than most enterprises have planned, because the bill includes the debt, not only the new tooling.

Trend 5. AI and cyber resilience redefine the CISO’s remit

Ninety-three percent of board directors view cyber risk as a threat to shareholder value. In addition, 98% expect it to increase over the next two years. This is where February’s regulatory volatility trend went. Gartner’s CISO Agenda for 2027 (G00859134, September 17, 2026) reports that 40% to 70% of CISOs now hold accountability for six adjacent areas, depending on the activity. Those areas are third-party risk, AI governance, disaster recovery, crisis management, business continuity and fraud prevention. Nearly half experienced burnout in the past 12 months, and nearly 60% say their team members did.

The same document counts the cost of AI vendor promises. Across 13 AI use cases in cybersecurity, only 39% of CISOs report significant ROI. In fact, 55% say vendor overpromise forced costly rework. Even so, boards are deliberately expanding their tolerance for technology risk to pursue AI growth. That is the finding of Gartner’s CISO Agenda Priority: Align Cybersecurity Investment With Board Risk Appetite (G00862663). The 2027 job is to measure business consequences rather than incident counts and to fund resilience for the risk the board has already accepted. Security consulting grows 7.8% to $31.7 billion in 2027 and professional services 8.5% to $34.5 billion. Those are the two slowest-growing of the three services subsegments in the forecast. Organizations are buying less outside strategy help as a share of spend. Instead, they are buying more software.

Trend 6. Agentic AI demands program oversight

Forty-six percent of senior cybersecurity professionals have observed production use of standalone SaaS GenAI agent platforms. Sixty-seven percent have also observed unsanctioned, employee-driven agent platforms through informal reports or monitoring. Meanwhile, 32% of IT workers using GenAI at work hide it. Only 16% of organizations have enforced workflows that govern employee use of third-party AI applications. That leaves what Gartner calls rogue automation running with excessive privilege. Despite the gaps, about 41% of GenAI prototypes reach production.

The Theme document reframes the whole category. Cybersecurity incidents are caused by AI agents taking ungoverned actions, not by how intelligent the model is. Gartner tells executives to treat every AI agent as an insider risk and to adopt risk frameworks built for nondeterministic systems. It also tells them to ban what it calls governance theater, meaning committees that can flag concerns but cannot block a launch. The Planning Guide’s version is to secure third-party agents first and extend application security to AI applications. Zero trust then becomes the architecture for agentic access. Zero trust network access grows 22.3% to $3.7 billion in 2027, the third-fastest category in the forecast. I covered the governance side in Gartner’s 2026 AI governance Hype Cycle and the testing side in the secure AI Hype Cycle.

Theme 3. Normalize AI adoption

Trend 7. Frontier AI models transform vulnerability management

New to the list, and the one most likely to change SOC metrics in 2027. Gartner’s argument is that frontier models created a supply-side shock. Vendors and open-source projects now use AI to find latent flaws in their own code. Consequently, the volume and frequency of urgent updates has jumped. The bottleneck is not patching capacity. Instead, it is testing and change management. Mandates to patch everything immediately therefore transfer risk into insufficient testing and self-inflicted downtime. Gartner calls patch speed a vanity metric and replaces it with maximum tolerable exposure time. I covered that shift in October in Gartner says frontier AI makes patch speed a vanity metric.

Maturity is still low. Fifty-two percent of organizations prioritize purely on CVSS ratings rather than active exploitability. Sixty-one percent still rely on traditional penetration testing, and only 39% use simulation technologies such as breach and attack simulation to confirm exploitability. The Planning Guide’s 2025 Impact of Emerging Tech on Infrastructure Security Operations Survey (n=249) finds only 48% weigh likelihood of exploitation and business impact when prioritizing. Similarly, only 41% use documented cross-team approval workflows to act on findings. Gartner’s Top Strategic Technology Trends for 2027 list adds autonomous exposure remediation. It predicts that by 2031 adopters will cut containment windows from weeks to within 48 hours. Threat intelligence spending grows 20.2% to $3.7 billion in 2027.

Trend 8. AI breaks traditional cybersecurity awareness tactics

Sixty percent of all data breaches involve the human element. Eighty-six percent of organizations also have standalone public SaaS GenAI in production. Over 57% of employees use personal GenAI accounts for work, and 33% admit inputting sensitive company information into them. The numbers from February have not improved. The threat side got worse. AI-assisted malicious emails, for instance, doubled in two years. Forty-one percent of organizations experienced at least one deepfake incident during an audio call with an employee. Likewise, 35% encountered one on a video call.

Gartner’s answer is to retire generic awareness training in favor of adaptive behavioral programs with AI-specific tasks. It also wants sanctioned GenAI tools to be easier to reach than unsanctioned ones. Cloud access security brokers grow 25.0% to $3.5 billion in 2027. CASB is the control most often used to see and govern SaaS GenAI usage. Enterprise data loss prevention grows 11.1% too. Those two lines are the budget expression of this trend.

Trend 9. AI-driven SOC solutions destabilize operational norms

Fifty percent of CISOs strongly agree that workforce resistance or talent gaps have slowed adoption of AI in security. The same gaps, they say, have cut the value of those investments. Similarly, only 17% of executive and AI leaders believe their organizations have enough talent to support AI initiatives. Inside the SOC, 26% have trained analysts to interpret enriched signals, and 35% have a structured skills transformation program. Meanwhile, 51% operate without a dedicated detection engineering function, which leaves them dependent on generic vendor rules.

Gartner’s forecast for the SOC workforce is the opposite of the headcount-cut narrative. By 2028, security operations teams will employ 50% more engineers than traditional analysts. By 2030 the SOC’s primary output will be validated cases with business impact and documented rationale rather than alerts reviewed. The Planning Guide also warns that AI SOC’s long-term value is unproven and that overpromising is the norm. Managed detection and response grows 9.8% to $12.6 billion in 2027, while managed security operations grows 10.7% to $17.3 billion. Organizations are still buying SOC capacity faster than they can build it.

What changed from the February list

Table graphic mapping Gartner's six February 2026 cybersecurity trends and one 1H26 trend to the 2H26 list, with status pills showing five kept, two removed and three new trends and a column describing where each trend went in the 2027 analysis
Five of February’s six trends survive into the 2H26 list, regulatory volatility is absorbed into the CISO remit trend, collaborative data security governance drops from the full 1H26 report, and three trends are new. Trend mapping by softwarestrategiesblog.com from Gartner’s Top Trends in Cybersecurity, 2H26 (G00860263) and the 2027 Planning Guide for Cybersecurity (G00856222). (Please click on the image to expand for easier reading)

Where the 2027 dollars move

Map the nine trends onto the forecast and a pattern shows up. The categories growing fastest in 2027 are the ones that reduce the attack surface AI agents and cloud workloads create. For example, cloud security posture management grows 29.1% to $8.1 billion. CASB grows 25.0%. ZTNA grows 22.3%. Cloud workload protection grows 21.8% to $9.2 billion. Threat intelligence grows 20.2%. Conversely, the categories shrinking are the legacy network controls those trends replace. Network access control falls 12.9% and intrusion detection and prevention systems fall 7.4%.

Horizontal bar chart of 2026 to 2027 growth rates for twelve Gartner security categories against the 11.6% market rate, from CSPM at 29.1% to network access control at minus 12.9%, each labeled with the 2027 trend it funds
The nine fastest-growing security categories in 2027 add $11.6 billion and fund attack surface reduction, agent access control and exposure management. The three slowest are human authentication and the legacy network controls those trends replace; NAC and IDPS lose $0.18 billion between them. Source: Gartner, Forecast: Information Security, Worldwide, 2024-2030, 3Q26 (G00862059). Growth computed and trend mapping by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

The Planning Guide’s fourth 2027 trend is attack surface reduction as a core part of defense in depth. It is the quietest of the four and also the best funded. Its planning considerations are endpoint hardening baselines and software supply chain controls. They also include CNAPP for the cloud attack surface, and DSPM plus AI security posture management for data overexposure. Endpoint protection platforms add $2.9 billion in 2027. That is the largest absolute increase of any category outside the other security software bucket where securing-AI tools sit.

What this means for CISOs planning 2027

Six of the nine trends describe the same failure from different angles. AI agents, the models behind them and the identities they borrow are operating inside enterprises faster than governance, identity, testing and training can adapt. Gartner’s Theme document gives that failure a name, the AI-Cyber Capability Gap, and a rule. You cannot scale AI value without scaling AI security first. The other three trends, quantum, technical debt and the CISO’s remit, are the structural debts coming due at the same time.

Diagram grouping Gartner's nine 2H26 cybersecurity trends into a four-step control chain for AI agents (verify components, bind identity, govern actions, validate exposure), two workforce trends and three structural debts, each card pairing a Gartner readiness statistic with the 2027 forecast line that funds it
Four of the nine trends form a control chain for AI agents, two describe the people who run it, and three are structural debts coming due at the same time. Each card pairs Gartner’s readiness figure with the 2027 forecast line that funds the response. Sources: Gartner, Top Trends in Cybersecurity, 2H26 (G00860263); 2027 Planning Guide for Cybersecurity (G00856222); Theme: Cybersecurity at the AI Frontier (G00862284); Forecast: Information Security, Worldwide, 2024-2030, 3Q26 (G00862059). Grouping and sequencing by softwarestrategiesblog.com. (Please click on the image to expand for easier reading)

Four moves for the 2027 plan

Four moves follow from the documents, in the order Gartner’s own numbers argue for.

Give every AI agent an identity you can trace to a human. Thirty-four percent of organizations run agents on human identity tools and 15% have no approach. Gartner’s 25%-of-breaches-by-2028 prediction is the cost of leaving that alone. Start with the agents carrying the most privilege and the least oversight.

Pilot PQC in 2027 or budget for triple the migration cost. Fifty-one percent have still not begun. Gartner’s 200% penalty applies to organizations that wait past 2027 to start piloting. The firewall refresh cycle already in the 3Q26 forecast is the first line item most CISOs will see.

Replace patch speed with tolerable exposure time. Fifty-two percent prioritize on CVSS alone. Frontier models will keep raising the volume of urgent fixes. The organizations that survive it will be the ones that validate exploitability first and then negotiate exposure-based SLAs with product teams.

Fund detection engineering before you fund AI SOC agents. Fifty-one percent of SOCs have no detection engineering function. Gartner’s 2028 forecast of 50% more engineers than analysts is a hiring plan, not a headcount cut. AI SOC agents, after all, amplify whatever the team already knows how to do.

Gartner’s six trends in February explained where $244.2 billion was going. The nine trends CISOs carry into 2027 explain where $276.1 billion goes. The fastest-growing lines in that forecast already match the trends with the lowest readiness scores. In short, the money is moving toward the gaps. The question for 2027 is whether the governance and the people move with it.

 

Gartner’s secure AI Hype Cycle predicts twice the incidents without preproduction offensive GenAI testing

Secure AI Hype Cycle overview. Gartner forecasts twice as many incidents by 2028 without preproduction testing as among proactive users. Of 24 technologies, one is under two years from mainstream adoption, 17 are two to five years away and six are five to ten years away.
Gartner, Hype Cycle for Secure AI, 2026 (G00846820), October 12, 2026. Incident comparison is a forecast. Technology counts and shares are my calculations from the complete Priority Matrix. Analysis and graphic by Software Strategies Blog. Click to open full size.

Gartner’s secure AI Hype Cycle predicts that enterprises skipping preproduction offensive testing of generative AI deployments will face twice as many cybersecurity incidents by 2028 as proactive users. That is a forecast, not a measured outcome.

Funding is already a constraint. Among organizations piloting or implementing offensive AI security technologies, 50% have either not secured funding or diverted money from existing cybersecurity initiatives. This finding describes surveyed adopters, not half of all enterprises.

Meanwhile, my count of the complete Priority Matrix puts 23 of 24 technologies at least two years from mainstream adoption. Security teams therefore need to control deployed agents before most specialist markets mature. The immediate questions are what each agent can do, which controls enforce its limits and what a new product would add.

Gartner’s Hype Cycle for Secure AI, 2026 (G00846820), published October 12, 2026, separates controls for employee AI use, enterprise-built applications and agents acting across systems. Those distinctions make it useful for testing a security architecture, not simply compiling a vendor shortlist.

Gartner’s original Hype Cycle for Secure AI, 2026, with all technology labels, adoption markers and phase labels.
Gartner, Hype Cycle for Secure AI, 2026 (G00846820), October 12, 2026, Figure 1, page 6. © 2026 Gartner, Inc. Click to open the full-size graphic.

Read more

20 Cybersecurity Startups to Watch in 2027

Bar chart ranking the latest verified funding round of each of the 20 cybersecurity startups to watch in 2027, from Island and Cyera at $400 million to Opal Security at $23 million, colored by the six sections of the article
Eighteen of the 20 closed their latest round in 2026. Bars show announced rounds, not total funding or valuation. Click the chart to open it full size.

Cyera paid $1 billion for Oasis Security and closed the deal on Sept. 3, 2026, 37 days after Globes reported the agreement on July 28. In late June, Oasis had made CRN’s list of the 10 hottest cybersecurity startups of 2026 so far. Startup lists go stale fast. Notable Capital’s Rising in Cyber 2026 names 30 companies picked by 150 CISOs and security executives, with funding data as of Q1 2026. A CISO setting a 2027 budget needs something narrower from both lists. Which young vendors produced evidence this year that they will still be independent, shipping and staffed when the contract renews?

Dot chart of each company's latest 2026 funding round by announcement date, showing 14 of the 20 rounds landing after March 31 and 12 after CRN's June 26 list, with markers for Notable Capital's Q1 data, the CRN list and Cyera completing its Oasis acquisition on Sept. 3
Each dot is a company’s latest verified round, placed on its announcement date. Keycard and 7AI raised in late 2025 and are not shown.

That question picked these 20. Every company here produced at least one verified 2026 event beyond a valuation, whether a priced round, a named production customer, a shipped product or a public technical result. Valuation earned no one a spot. Eleven of the 20 appear on neither the CRN nor the Notable list.

Eighteen of the 20 closed their latest round in 2026. That money turns into hires, product launches and sales coverage next year, when buyers will feel it. Several companies here have already tied specific 2027 deliverables to the cash.

The research cutoff is Oct. 17, 2026. To qualify, a company had to be privately held and independent on that date. Late-stage companies qualify under the same rule as seed-stage ones, so Cyera and Island stay in. A completed acquisition disqualifies a company. So does an announced sale, which is why Oasis is out.

Securing AI agents company profiles for Onyx Security, HiddenLayer, Reco and Harmonic Security, showing each company's latest verified round, total funding where disclosed, buyer, 2027 watch item and what to validate
Securing AI agents at a glance. Bars share one scale across all six sections.

Read more

Gartner’s $9.76T IT spending forecast: servers will outspend all IT services by 2030

Gartner IT spending forecast 2024 to 2030 rising from $5.04 trillion to $9.76 trillion, stacked by data center systems, software, IaaS, IT services, communications services and devices, with six KPI tiles
Gartner’s 3Q26 worldwide IT spending forecast by segment, 2024-2030, with six tiles on the numbers that matter most. Click any chart to open it full size. Source: Gartner (G00862058, Sept. 25, 2026), analysis by softwarestrategiesblog.com.

Gartner’s latest IT spending forecast puts worldwide IT spending at $6.39 trillion in 2026 and $9.76 trillion by 2030. Spending grows 13.2% in 2026 in constant currency, or 14.5% in current dollars, and compounds at 11.2% a year from 2025 to 2030.

One subsegment drives that growth more than any other. Server spending grows from $386.5 billion in 2025 to $1.97 trillion in 2030. That $1.58 trillion gain is 38% of the $4.18 trillion the IT market adds over the period. By 2030, the world spends more on servers than on all IT services combined.

Gartner calls building AI compute capacity “the largest infrastructure build-out in history.” This edition makes the build-out bigger. Gartner raised its 2030 total by $816.8 billion from last quarter, and data center systems account for $794.9 billion of that increase.

I built this analysis from Gartner Market Databook, 3Q26 (G00862058), published September 25, 2026. Segment totals, growth rates and revisions come from the report. Subsegment and regional figures come from the 7,896-row dataset Gartner publishes with it, which covers 19 subsegments, nine regions and 47 countries. The shares, dollars added and comparisons are my calculations.
Read more

Gartner’s $812.6B app software forecast sends 38% of new spending to CRM as agentic AI targets $116B of legacy apps

Gartner enterprise application software forecast 2024 to 2030 rising from $384.8 billion to $812.6 billion, stacked by CRM, ERP, other application software, email and authoring, analytic platforms, supply chain management, content services and project and portfolio management, with six KPI tiles
Gartner’s 3Q26 enterprise application software forecast by market, 2024-2030, with six tiles on the numbers that matter most. Click any chart to open it full size. Source: Gartner (G00862072, Sept. 22, 2026), analysis by softwarestrategiesblog.com.

Enterprise application software spending reaches $496.3 billion in 2026 and $812.6 billion by 2030, according to the latest Gartner application software forecast. That is a 12.8% compound annual growth rate in constant currency, revised up from 12.7% last quarter.

The headline growth rate hides a sharper story. CRM alone captures $119.2 billion, or 38%, of the $316.2 billion the market adds between 2026 and 2030. In the same report, Gartner says cross-functional agentic AI will cannibalize $116 billion from traditional application software spending by 2030.

Those two numbers are close in size, and they point to the same conclusion. Most of the new money in application software goes to the markets where agentic AI is proving value first. Legacy suites that do not embed agents compete for a shrinking slice.

I built this analysis from Gartner’s Forecast: Enterprise Application Software, Worldwide, 2024-2030, 3Q26 (G00862072), published September 22, 2026. All market, subsegment and regional figures come from the 7,567-row dataset Gartner publishes with the report. The shares, dollars added and comparisons are my calculations.
Read more

Gartner’s $239B AI forecast: Agentic workflows take half of GenAI model revenue

Detailed 2024–2030 GenAI revenue chart with annual spending, agentic shares and dollar allocations; intermediate agentic shares are explicitly labeled SSB scenarios.
Gartner publishes the agentic share of GenAI model revenue for 2025 (5%) and 2030 (50%). The 2024 and 2026 to 2029 bars are my estimates and are labeled SSB. By 2030, agentic revenue reaches about $69.6 billion of $139.2 billion. Source: Gartner, G00855897 and G00861842. Chart and analysis by softwarestrategiesblog.com.

Agentic workflows will drive 50% of generative AI model revenue by 2030, up from 5% in 2025. Gartner published that projection on September 17, 2026, in its forecast analysis of the generative AI models market (G00861842). The number changes how every enterprise buyer should read the firm’s $239 billion AI platforms and models forecast.

Multistep reasoning, tool integration, and repeated validation steps multiply the inference events behind every completed business process. Gartner puts the impact at $38 billion in additional spending by 2030 from scaling agentic workflows alone.

Falling inference prices won’t offset it. Cheaper units unlock deeper automation, and deeper automation drives up token volume per task faster than prices fall. Gartner calls this a structural tailwind for GenAI model spending.

That consumption flywheel sits inside a market growing from $39 billion in 2025 to $239 billion by 2030 at a 42.8% compound rate, per Gartner’s June 25 forecast (G00855897). I built this analysis from both Gartner reports and Gartner’s country-level dataset, which covers 1,316 rows across 9 regions and 7 years. For the agentic spending crossover that sets up this post, see Gartner’s $5.95 trillion AI forecast puts the chatbot era on a 2027 deadline (August 5, 2026).

Model revenue is also one of the streams that has to pay for the infrastructure build-out. How far it falls short is the subject of my AI revenue gap comparison across Gartner, Goldman, Bain and JPMorgan.
Read more

Gartner’s AI security forecast exposes 162x services growth that still trails software 2 to 1 in new spending

Gartner AI-amplified security forecast, growth multiple versus net-new dollars by segment, 2024 to 2030

Growth multiple versus net-new dollars added, 2024 to 2030. The segment ranking inverts between the two measures.

Every chart and table in this analysis opens full size when you click it.

Security services inside Gartner’s AI-amplified security market were worth $364 million in 2024. Gartner now projects $59 billion by 2030. That is 162 times larger in six years, compounding at 133.5% annually, the steepest curve anywhere in the forecast. I have tracked this forecast through every quarterly update, and no segment has ever moved like this one.

What Is AI-Amplified Security?

Gartner’s term for the share of existing security spending flowing to products with AI built in. Endpoint protection, firewalls, identity, and network security that now embed AI-driven detection, autonomous remediation, and agent-based response. Not a new category. Existing budgets redirecting toward AI-native capabilities. The companion forecast for securing AI itself reaches $16.4 billion in 2030. Gartner publicly confirms the 75%-by-2028 adoption projection.

The share table and the dollar table tell different stories. Software’s share of this market falls from 87.1% to 61.3%, and services picks up almost every point software gives up. Read only that and you conclude software is losing. Run the arithmetic on net-new spending and software still collects $116.5 billion of the $194.4 billion the market adds between 2024 and 2030. That is 60 cents of every new dollar, and it puts software ahead of services 2 to 1 on net-new spending.

Both things are true at once, and the gap between them is where security budgets get set wrong.

The numbers come from Gartner’s Forecast Analysis: AI-Amplified Security, Worldwide, 2026 (G00846160, August 4, 2026) by Shailendra Upadhyay. It is the first time Gartner has split AI-amplified security into software, services, and network security across a full seven-year window. The totals reconcile cleanly with the 2Q26 AI spending forecast, which carried AI-amplified security at $204.5 billion in 2030 without breaking out the segments underneath it.

Read more

34% of enterprises govern AI with policies they only partly follow. Gartner’s first AI Governance Hype Cycle shows CISOs what to fund first.

Gartner Hype Cycle for AI Governance, 2026, plotting AI governance platforms, AI TRiSM, AI agent identity and responsible AI across the innovation trigger, peak, trough and slope stages

Gartner Hype Cycle for AI Governance, 2026, showing AI cybersecurity governance, AI governance platforms, and agentic AI risk innovations plotted across the innovation trigger, peak, trough, slope, and plateau phases. Please click on the graphic to expand for easier reading.

Gartner’s 2026 AI Leaders Effectiveness Survey found that 34% of organizations have well-defined AI governance structures and policies for managing risks, ethics, and compliance. Those same organizations report only partial adherence to the rules they wrote, while another 22% still rely on basic or ad hoc policies. Only 7% qualify as recognized leaders in ethical AI.

That gap between writing the policy and living by it is what Gartner’s first Hype Cycle for AI Governance, 2026 is built to address. Published July 21, 2026, the inaugural cycle plots 32 innovations and ranks each by benefit rating and years to mainstream adoption. CISOs and enterprise architects get a planning map that builds on Gartner’s forecast that agentic AI will overtake chatbot spending by 2027. Throughout this post, I use “rogue agents” as my editorial shorthand for unsanctioned AI agents operating outside governance perimeters.

The two clusters that carry the security agenda

The report organizes its 32 innovations around six enterprise trends. Two carry the security agenda. Agentic AI oversight and life cycle governance is the first, grouping agentic AI governance, agent development life cycle, AI agent identity, AI engineering, AI gateways and AI governance platforms under one trend. Advancing AI security is the second, covering AI TRiSM, AI cybersecurity governance, disinformation security, zero-trust data governance, mechanistic interpretability and AI product attribution and transparency. Together they define where governance stacks connect to identity systems, traffic controls and incident response playbooks. For how these gaps show up in spending data, see my analysis of Gartner’s $248.9B security forecast.

Five innovations that pay off in under two years

Table 1, the Priority Matrix, ranks every innovation by benefit rating and adoption timeline. Only five entries land in the “Less Than 2 Years” column. Responsible AI sits alone in the Transformational row, Gartner’s highest rating. AI guardrails, data access governance, digital ethics and ontologies all carry High benefit ratings at the same timeline. Gartner calls these the near-term priorities for scalable governance. CISOs should fund them first.

Table 1: Gartner Priority Matrix for AI Governance, 2026, mapping transformational, high, moderate and low benefit innovations against years to mainstream adoption

Please click on the image to expand for easier reading.

The two-to-five-year column is the densest band. Thirteen innovations carry a High benefit rating there, including agentic AI governance, AI agent identity, AI TRiSM, zero-trust data governance and third-party risk management. That band is where CISOs will build governance stacks over the next several budget cycles. For context on spending already flowing to AI-related capabilities, see my breakdown of Gartner’s $244.2B security forecast.

Where AI cybersecurity governance actually lands

AI cybersecurity governance carries a Moderate benefit rating at the Innovation Trigger, five to ten years from mainstream adoption. That placement may surprise CISOs who expected Gartner to rate it higher. The profile’s key goals are to prevent shadow AI, minimize attack surfaces and ensure visibility and response to incidents. The placement is a market signal, not a dismissal. Tooling is early, but the need is urgent enough that CISOs should treat AI cybersecurity governance as an architecture requirement today.

What the Priority Matrix tells your board

The report leads with two strategic planning assumptions that carry board-level weight. Enterprises implementing AI governance will outperform ungoverned competitors in AI adoption by 25% by 2029. The downside is just as concrete. Autonomous agents identifying minor consumer rights violations and turning them into lawsuits will increase corporate settlement costs by 15% over the same period. Growth-oriented executives respond to the first number. Risk-averse ones respond to the second.

Turning the curve into controls

Start by mapping your AI agent footprint, sanctioned and unsanctioned, across SaaS platforms, internal applications and shadow IT. From there, match the Hype Cycle’s innovations to four governance domains. The identity and access layer runs on AI agent identity and AI governance platforms. Data classification draws on AI guardrails, data access governance and zero-trust data governance, while AI gateways and AI TRiSM handle traffic mediation. Agent risk management anchors in agentic AI governance and the agent development life cycle. Build all four as shared services. For how agent sprawl is reshaping security spending, see my roundup of agentic AI forecasts and market estimates, 2026.

What these numbers add up to

Gartner’s inaugural Hype Cycle for AI Governance puts 32 innovations on the curve. Only five reach mainstream adoption in under two years. The security-relevant capabilities cluster in the two-to-five-year band, which means CISOs have a narrow window to build governance stacks before agent footprints outpace controls. The 34% adherence stat is the warning, and the Priority Matrix is the roadmap out of it.

This post is my personal reflection on Gartner’s AI governance research from a CISO and enterprise architecture perspective. It does not represent any employer or client.

Source: Gartner, Hype Cycle for AI Governance, 2026, Svetlana Sicular, Var Shankar, Lauren Kornutick, Sumit Agarwal, 21 July 2026, G00854164.

For each month’s AI agent attacks, exploited CVEs and breaches with primary sources, see my monthly AI security news briefing.

Gartner’s $5.95 trillion AI forecast puts the chatbot era on a 2027 deadline

Chart from Gartner's 2026 AI spending forecast: embedded chatbot spending peaks at $272.6B in 2027 then declines to $205.8B by 2030 while embedded agentic AI rises to $1,003.6B

Spending on the chatbots and assistants embedded in enterprise software peaks at $272.6 billion in 2027 and then shrinks every year through 2030. Gartner buried that projection inside the 2Q26 update of its worldwide AI spending forecast, published July 24, and it matters more than the headline total. By 2030, embedded chatbot spending falls back to $205.8 billion, a hair above its 2025 starting point.

Embedded agenticAI takes the money instead, growing from $88.2 billion in 2025 to $1 trillion by 2030, an 11.4x expansion inside a single software category.

None of that slows the topline. Worldwide AI spending reaches $2.67 trillion in 2026, up 49.5% from 2025, on its way to $5.95 trillion by 2030. The figure Gartner published in May was $2.59 trillion for this year. Ninety days later, the client-facing number runs $74.8 billion higher, and the firm added $496.8 billion to its comparable 2025 through 2030 outlook in a single quarter.

Four tables below show where the money lands, which segments stall, and what Gartner changed its mind about between April and July.

Where $5.95 trillion lands

Table: worldwide AI spending by market, 2025 to 2030, in billions of dollars, from AI infrastructure at $2,792.9B to AI data at $20.7B, grand total $5,950.6B with 27.2% CAGR, Gartner July 2026

Infrastructure stays the biggest line through 2030 at $2.79 trillion, even as its share of total spending slides from 55% in 2025 to 46.9% at the end of the window. AI-optimized servers alone reach $981.7 billion by 2030, a 3.4x jump from 2025, and AI processing semiconductors add another $656.4 billion. O

One caution before quoting the total anywhere. Gartner’s note flags the forecast as a view across the whole AI value chain, so the chip and the server it ships inside both get counted. Read $5.95 trillion as the size of the AI economy, not as net end-user budgets.

Devices carry more of the infrastructure number than most readers expect. Business and consumer AI devices combine for $904.4 billion in 2030, and $647.4 billion of that is consumer hardware, the AI PCs and phones landing in shopping carts rather than data centers.

Growth flattens fast after next year. Total spending rises 49% in 2026 and 36% in 2027, then steps down to 21%, 18% and 15% through 2030, while infrastructure decelerates from 51% growth this year to 9% at the end of the forecast.

Of the $883.8 billion in net-new AI spending arriving in 2026, infrastructure absorbs $502.5 billion, or 57 cents of every new dollar. The shape of the curve says the buildout peaks now and software inherits the growth.

Farther down the board, AI cybersecurity at $220.9 billion and AI agents and assistants at $219.9 billion finish 2030 within $1 billion of each other. Gartner created the agents category only this quarter.

Agentic AI crosses $1 trillion inside enterprise software

Table: chatbots versus agentic AI worldwide spending, 2025 to 2030, showing embedded agentic AI growing 62.7% a year to $1,003.6B while embedded chatbots grow 0.6%, Gartner July 2026

Gartner rebuilt its segmentation this quarter, splitting cross-functional and consumer agents out of AI software and adding consumer agents to the forecast for the first time. The new structure exposes a replacement cycle the old rollup hid. Inside enterprise software, agentic AI overtakes chatbots in 2027, the same year chatbot spending tops out, and from that peak to 2030 the embedded chatbot line surrenders $66.8 billion.

Cross-functional agents, the ones that work across software from multiple vendors, start from a base of zero. Gartner books $347 million for cross-functional agentic AI in 2026 and $78.1 billion in 2030, a number it raised this quarter on the thesis that these agents begin cannibalizing traditional SaaS by decade’s end. The ceiling matters as much as the curve. Against $1.21 trillion in total 2030 AI software spending, $78.1 billion says the incumbents hold the decade, because data access, integration complexity and execution reliability hold the category back from serious SaaS competition until 2030, in Gartner’s read.

The buy-versus-build verdict is just as lopsided. Agent builder platforms, the tooling for constructing your own agents, reach only $12.6 billion by 2030, so embedded agentic AI outspends them nearly 80 to 1. The first production agent most companies run will ship inside software they already own. Gartner describes exactly that race, with vendors across software categories embedding agentic AI to defend their installed bases against cross-functional challengers.

Consumer agents barely register yet in dollar terms. Gartner carries $26.9 million for consumer agentic AI in 2026, then $17.7 billion in 2027 as paid consumer agents arrive at scale, building to $51.8 billion by 2030. Adding consumer agents and assistants lifted Gartner’s 2030 total by $133 billion. For how these agent numbers stack against other analyst estimates, see my roundup of agentic AI forecasts and market estimates, 2026.

AI security expands 8.5x and splits in two

AI cybersecurity grows from $25.9 billion in 2025 to $220.9 billion in 2030, an 8.5x expansion at a 53.5% compound rate. Spending in the category grew 140% in 2025, and Gartner models another 98% jump this year. AI cybersecurity and AI data were also the only two markets left completely untouched between the 1Q26 and 2Q26 forecasts, which makes security the steadiest conviction in the entire model. For the standalone security spending outlook, see my breakdown of Gartner’s 2Q26 information security forecast.

Two markets move at different speeds inside the category. AI-amplified security, meaning AI capability inside security tooling, carries the volume and reaches $204.5 billion by 2030. Securing AI, the discipline of protecting AI systems themselves, runs smaller and faster, from $1.5 billion in 2025 to $16.4 billion in 2030 at a 60.4% compound rate.

Set the security numbers against the agent forecast and an exposure gap opens. The 2030 outlook has enterprises running $1.08 trillion of embedded and cross-functional agentic software while spending $16.4 billion to secure AI systems, roughly $66 of agentic software for every $1 of securing-AI budget. AI observability and governance tooling adds just $3.9 billion more. A software wave that large riding on a security ratio that thin is the budget argument CISOs should be starting now.

The fastest growth goes to whatever cuts the bill

Table: the 10 fastest-growing AI segments 2025 to 2030, led by synthetic data generation at 142.5% CAGR and DSLMs at 84.5%, with securing AI at 60.4%, Gartner July 2026

Rank every segment by compound growth and one pattern jumps out, because the fastest-growing lines are the ones that make AI cheaper. Synthetic data generation leads the entire forecast at a 142.5% compound rate, expanding 84x from $146 million in 2025 to $12.2 billion in 2030. AI-ready datasets, the licensed real-data alternative, peak at $583 million in 2029 and then decline, leaving synthetic data outselling licensed data 22 to 1 by 2030. Gartner is forecasting the substitution of purchased data itself.

Domain-specific language models tell the same cost story at larger scale. DSLMs and specialized models grow 210% in 2026 and compound at 84.5% through 2030, rising from 12.2% of all model spend to 24.3%. Cost pressure also explains the strangest revision in the update. Gartner cut $57.8 billion in cumulative dollars from its generative AI model forecast while raising the segment’s 2026 growth rate from 110% to 117%, which nets out to more deployments running on cheaper models and smaller checks. Arunasree Cheparthi, a senior principal research analyst at Gartner and one of the forecast’s authors, said in the firm’s July 20 platforms and models announcement that spending “is shifting toward providers who can demonstrate clear value.”

What Gartner changed in 90 days

Table: what Gartner changed between its 1Q26 and 2Q26 AI spending forecasts, with AI infrastructure revised up $136.0B for 2030 and AI services revised down $11.2B

Between the April forecast and this one, Gartner added $500.8 billion to AI infrastructure across the 2025 through 2030 window, the largest revision in the update, and did it while flagging memory-related price increases. The note calls infrastructure demand inelastic to that pricing pressure, because hyperscalers keep buying AI-optimized servers on the conviction that model capabilities improve through 2030.

Gartner’s September consolidation research asks whether that infrastructure pays back. I compare its model with Goldman, Bain and J.P. Morgan in AI revenue gap forecasts compared; by my calculation the shortfall in Gartner’s model exceeds $6 trillion through 2032.

Software took the other side of the trade. AI software gained $191.9 billion and application development platforms picked up $10.7 billion. Gartner lifted the 2026 app-dev growth rate from 28% to 39% as enterprises build custom AI applications and demand usage tracking to prove the spend. The cuts land on everything that resembles consulting or plumbing. AI services lost $80.7 billion across the window, with every single year revised down, and platforms for data science and machine learning lost $68.2 billion, including an 8% cut to 2027 alone.

The services cut hides a structural shift rather than a retreat. Gartner still sizes AI services at $1.25 trillion in 2030, but the growth belongs to indirect services, which compound at 34.5% and pass direct, consulting-led engagements in 2028. Direct AI services compound at 15.7%, less than half the indirect rate. Buyers are routing transformation budgets through software and cloud purchases instead of billable hours.

Three dates to plan against

2027 is the year chatbot spending tops out and agentic AI takes over inside enterprise software, which gives any vendor still selling assistant-branded features through the end of next year to ride what growth remains. By 2028, indirect services pass consulting-led engagements and infrastructure growth drops to 15%, so the buildout stops flattering everyone’s numbers. And 2030 arrives with $1.08 trillion of agentic software guarded by $16.4 billion of securing-AI spend. The first two dates decide where the money goes, and the third decides what happens when it arrives unprotected.

This post is my personal reflection on Gartner’s AI spending research from an industry analyst perspective. It does not represent my employer.

Source: Gartner, Forecast: AI Spending, Worldwide, 2025-2030, 2Q26, Kay Arnott, Jon Erensen, Amarendra, Adrian O’Connell, Arunasree Cheparthi, Naresh Singh, Peter Middleton, Hardeep Singh, Shailendra Upadhyay, Rishi Padhi, 24 July 2026, G00855896.