Model

Alphabet plans up to $205B to speed up AI the same month it was sued for slowing down

Alphabet's new DeepMind chief spent his first public week promising Gemini 4 early, while his staffers took VC meetings, a safety engineer quit over the pace, and a subscriber suit alleged Google agreed with rivals to slow development. The same shareholders fund the speed-up and parts of the exodus.

Vincent JiangVincent Jiangthree-minute read
Share
Demis Hassabis, who led Google DeepMind until August 2026 and is now Alphabet's chief scientist
1 / 6Slide 1 of 6
Demis Hassabis handed day-to-day control of Google DeepMind to Koray Kavukcuoglu on 5 August and became Alphabet's chief scientist.

Speed was the new chief's first public sentence

Koray Kavukcuoglu's first public appearance as head of Google DeepMind was a speed pledge. On 23 September he said Gemini 4 has entered early post-training and is already powering Google's Antigravity coding tool internally, with a release he hopes comes "much earlier" than the end of 2026; no date was named 123. Gemini 3.5 Pro, promised for June, never arrived 3.

The market had priced the doubt first. GOOGL−1.04% — GOOGL, down 1.04 percent today fell 3.8% that day, the tail of a near-5% two-day slide on Meta's−4.82% — Meta, down 4.82 percent today Muse hardware push, then recovered 0.6% after hours on his words 1. GPT-5.5 is now the default for coding workloads, Claude Code the main alternative, and one Google employee's word for the company's models was "behind" 1.

August handed the lab to the accelerator camp

Demis Hassabis gave up day-to-day control on 5 August to become Alphabet's chief scientist, and Jeff Dean left after 27 years to found Discovery Loop, with Google as founding investor 4. Sundar Pichai framed the reshuffle as acceleration: "We have to accelerate all this work and stay focused on the AI frontier" 4.

Shareholders carry the invoice. Capex ran $32.3B in 2023 and $91.4B in 2025, and 2026 guidance sits at $195B to $205B 54. The money is buying growth: cloud revenue rose 82% year over year to $24.8B, on a $514B contract backlog 41.

Alphabet nearly tripled capex since 2023, and 2026 guidance doubles it again

  • Estimate
$0B$50B$100B$150B$200B$250BFY2021$24.64BFY2022$31.49BFY2023$32.25BFY2024$52.54BFY2025$91.45BFY2026 guide$195–205B
Data
ValueRange
FY2021$24.64B—
FY2022$31.49B—
FY2023$32.25B—
FY2024$52.54B—
FY2025$91.45B—
FY2026 guide (estimate)$205B$195–205B
Annual capital spending, USD billions. FY2021 to FY2025 as reported in Alphabet's SEC filings via Sharadar; the FY2026 bar is company guidance, with the guided low and high marked. Sources: Sharadar annual fundamentals; Gizmodo, 5 August 2026.4,5

The exits run in both directions

Fifteen current and former DeepMind staffers met this month at a five-star hotel across from Google's London headquarters, trading tips on what Silicon Valley will fund, per one attendee 67. The template pays: David Silver's Ineffable Intelligence banked a $1.1B seed in April at a $5B valuation, Europe's largest ever, then added four DeepMind veterans as cofounders 8.

Robert O'Callahan went the other way. The engineer, whose chip-design tools will make AI "much cheaper and faster," resigned on 24 September, calling the near-term pursuit of superintelligence "inherently irresponsible" and saying colleagues he raised it with share serious concerns, mostly in private 910.

The suit says Google promised the opposite

On 18 September, paying subscribers filed an antitrust class action in the Northern District of California alleging Google, OpenAI$1.18T — OpenAI, private, latest valuation $1.18T, Anthropic$350B — Anthropic, private, latest valuation $350B and xAI$200B — xAI, private, latest valuation $200B coordinated to slow frontier development; The Inference covered the filing on 20 September 11. It landed six days after Dario Amodei's essay urging labs to "pace the frontier" drew same-day backing from Sam Altman, Elon Musk and, with reservations, Hassabis 1112.

Plaintiffs want treble damages and a nationwide subscriber class. The claims are unproven, and Amodei's own essay flagged the antitrust exposure, asking the government for a narrow waiver to allow the coordination 1112.

The pace decides the next leg

Three markers: Gemini 4's actual ship date, any incorporation filing from the breakfast group, and the class-certification ruling. Google is being sold off for not shipping fast enough and sued over allegations it agreed to slow down, in the same month 11.

The bill and the benefit split cleanly. Shareholders and paying subscribers carry the cost, rivals hold the coding market Google ceded, and the leavers bank the seeds. Every camp here is betting on the same variable, the pace.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Loading GOOGL

Reader comments

0 comments

    Model

    HeyGen sells Avatar 6 while Tavus's Griffin scores 3.83 against a human's 3.92 and still has no price

    HeyGen put Avatar 6 on general availability on November 17 with a public API and first place in New Measure's blind Avatar Arena; Tavus's Griffin, which Tavus says is the highest score ever recorded for an AI on NVIDIA's face-to-face benchmark, is still a preview with no price. Revenue accrues to HeyGen today because Avatar 6 is the model on sale; Tavus's best model is not.

    Model

    Pika's Enact lets creators direct AI video mid-scene, with answers in under a second

    Enact, posted November 18, lets a creator drag an AI actor mid-scene, talk to the cast by name and redo the camera after locking a performance. Pika open-sourced the model underneath and kept the directing layer closed, a bet that value in AI video is moving from model quality to workflow.

    Model

    Anthropic falls 11% toward its $83 IPO price as Microsoft undercuts Claude

    Anthropic fell about 11% Friday morning, back toward the $83 IPO price, as Microsoft's price attack on Claude met the open market. Microsoft holds Anthropic's compute contract, its shares and a cheaper rival product; a leveraged income ETF on the two-day-old stock listed the same morning.

    Model

    Ramp's invoices show Anthropic held 58.7% of US business AI spend vs OpenAI's 37.2% before its record IPO

    Ramp's index of 85,000 US business bill-payers is the first independent check on the record $81.3 billion IPO: the enterprise lead over OpenAI is real, and so is the falling price of what Anthropic sells.

    Model

    Google's Gemini 4 Neon took the coding lead the day before Anthropic's $1.5 trillion debut

    Google's mid-tier model debuted at No. 1 on Artificial Analysis's coding-agent index at $6.86 a task against Claude Opus 5.5's $12.61, a day before Anthropic's Nasdaq debut. A $1.5 trillion valuation built on coding-led growth now meets a measured, cheaper rival, and the exposure sits with whoever paid $83.

    Model

    ElevenLabs, valued at $22 billion, puts its voice agents inside NICE, Genesys, Amazon Connect and Five9

    Salesforce and Five9 have put more than $3 billion into buying customer-service AI this month. ElevenLabs, valued at $22 billion after a September share sale, embedded its agents natively into Five9's platform and three rivals instead, and the evidence points to pricing power sitting with the channel, not the model maker.

    Subscribe to TheInference for free