Gartner lifts 2026 AI spending forecast to $2.7 trillion, up 49.5%
Two signals from the weekend, one about money and one about words: the AI spending forecast went up again, while the president of the United States ran a poll on what to call the technology.
Gartner raised its forecast for worldwide AI spending in 2026 to $2.7 trillion, a 49.5% increase on 2025 — up from the $2.59 trillion and 47% growth the firm published in May. AI infrastructure, which the analyst house defines as AI-optimized cloud plus data-center construction, remains the largest single category, and Gartner's John-David Lovelock describes the build-out as "the largest infrastructure project in human history," adding that demand has stayed inelastic even as memory prices climb. The revision is the headline. The more useful line sits underneath it: Lovelock says enterprises are commissioning fewer large-scale transformation programs and buying more small ones, mining the AI features already embedded in software they own — despite vendor lock-in, data sovereignty and cost-overrun risk. Gartner puts the combined services opportunity at $1.2 trillion by 2030. Read the two halves together and the market is not slowing, it is re-shaping: the infrastructure line belongs to hyperscalers and labs, while the enterprise side quietly moved from projects to features. We cited the earlier number when the release wave crested — Model fatigue is real: AI labs can't stop shipping.
Trump opened a public poll asking whether "artificial intelligence" should be renamed, offering three alternatives: "Superior Intelligence," "Extreme Intelligence" and "Supreme Intelligence." On Truth Social he wrote that the existing term is "inaccurate, and very ineloquent," then asked followers to vote on "the best name for this ever growing 'Revolution?'" Chinese outlets tracking the post put it past 180,000 votes with Superior Intelligence ahead. The naming stunt is thin on its own; the posture around it is not. In the same stretch he called AI safety warnings a "hoax" and said the only guardrail the technology needs is "a STRONG AND SMART (High IQ!) PRESIDENT," and he stood up the federal "AI Force" we covered over the weekend — Trump creates an 'AI Force' and says a czar will run it. Worth noting that two of his three suggestions abbreviate to the same two letters.
October's model calendar is filling up, anchored by a confirmed OpenAI DevDay on September 29 in San Francisco. Chinese tech press rounds up at least ten flagships expected around the holiday window: OpenAI's GPT-6 Sol, Terra and Luna (Sol is widely expected at DevDay, though OpenAI has not announced it), Anthropic's Opus 5.2 and Fable 5.2, Google's Gemini 4.0 Pro and Flash, xAI's Grok 4.7, and on the Chinese side Qwen4, GLM-5.5, MiniMax M3.1, Tencent's HY4 Pro and DeepSeek V4.1 Pro. None of those are announced products, and the parameter-count talk around DeepSeek and GLM — doubling past a trillion — is rumor, not spec sheet. What is real: StepFun's Step 5 Preview is already live, Gemini 4 Pro has been sitting anonymously on Arena for days, and DeepSeek V4.1 Flash already shipped and retired V4 Pro. The clustering is the story — labs stacking launches into a two-week window and letting leaks do the announcing.
What to watch: whether Gartner's embedded-AI line holds as that wave lands. Features inside software already sold are cheaper to buy than transformation projects, and the enterprise buyer has already moved.
If the money is shifting from AI projects to AI features, what does that mean for the labs selling frontier capability? Tell us in the comments.
Sources: Gartner — Worldwide AI spending to grow 49.5% in 2026 · Sina Tech (新浪科技) — 2026 global AI spending at $2.7 trillion · Economic Daily (經濟日報) — Gartner raises AI spending estimate · Anadolu Agency — Trump opens poll proposing renaming artificial intelligence · Ajel — Trump to create AI Force and launches renaming poll · OpenAI DevDay 2026 · 快科技 (Mydrivers) — a dozen flagships queued around the holiday window · NextBigFuture — Grok 4.7 and Fable 5.2 also in September
