Google's 'Frozen v2' Chip Aims to Supercharge Gemini AI
24 Jul 2026
Google is internally developing a new server chip, codenamed Frozen v2, designed to make its Gemini AI models run significantly more efficiently, according to a report from The Information later confirmed by TechCrunch. The news sent Alphabet's stock up roughly 3% on Monday morning.
What We Know
Alphabet is reportedly designing Frozen v2 as custom server silicon aimed at improving how efficiently Gemini models operate. The chip is expected to be released sometime in 2028 and could deliver between 6 and 10 times the efficiency of Google's current AI chips, measured by tokens generated per unit of power.
This emerges alongside Google's broader AI strategy, which reportedly includes planned spending of $180 billion to $190 billion. Google offered a measured public response, stating: "Our teams are constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers."
The Bigger Chip Race
Google isn't alone in pursuing custom silicon. The timeline shows a broader industry pattern:
- June 2026: OpenAI announced its first custom chip, an inference processor called Jalapeño.
- Early July 2026: Anthropic was reported to be discussing a chipmaking partnership with Samsung.
- July 20, 2026: TechCrunch's report on Frozen v2 followed The Information's original story, coinciding with Google's stock bump.
This suggests major AI labs are increasingly moving toward vertical integration — designing their own hardware rather than relying solely on third-party chipmakers.
What's Still Unclear
Several important details remain unconfirmed. There's no public information on Frozen v2's architecture or manufacturing partner, and it's unclear how the $180-190 billion spending figure breaks down between chip development, infrastructure, and other AI initiatives. Google has not confirmed the accuracy of the 6-10x efficiency claim or disclosed the methodology behind it. There's also no direct comparison available yet between Frozen v2 and competitor chips like OpenAI's Jalapeño. Additionally, it's unclear whether the stock increase was driven primarily by the Frozen v2 report or by other market factors.
Why Founders Should Care
For startups building on Gemini APIs, Frozen v2 could eventually translate into lower operating costs or improved performance — but this is likely years away and depends heavily on whether the efficiency claims hold up in practice. The chip race among Google, OpenAI, and Anthropic may signal a broader industry shift toward vertical integration, which founders relying on these platforms should probably monitor as a potential signal of platform dependency risk down the line.
Given the 2028 timeline, founders should not expect near-term infrastructure changes from Frozen v2 specifically. However, the scale of AI infrastructure spending being reported across major players may suggest continued investment momentum in AI tooling and infrastructure, which could plausibly shape the broader funding and product landscape for AI startups in the coming years.
Key Risks to Watch
Several factors could complicate this picture. The multi-year development timeline means competitors could ship efficiency gains sooner, and the 6-10x performance claim is unconfirmed and could shift or underdeliver by release. Heavy AI infrastructure spending in the $180-190 billion range increases financial exposure if efficiency gains don't materialize as projected. And with OpenAI and Anthropic pursuing their own chip strategies, any first-mover advantage Google might gain could erode quickly as the broader industry race intensifies.