Loading article…
Google’s new “Frozen v2” chip could be 6‑10× more power‑efficient, key to handling its $462 B cloud backlog and easing AI spend concerns.
Google’s internal “Frozen v2” server chip, slated for a 2028 release, is projected to be six‑to‑ten times more power‑efficient than its current AI silicon, a gain that could help the company service its $462 billion cloud backlog and calm investor worries over massive AI spend [1].
| At a glance | |
|---|---|
| Chip name | Frozen v2 |
| Efficiency gain | 6‑10× vs. existing AI chips |
| Expected release | 2028 |
| Cloud backlog | $462 billion |
Alphabet’s AI‑focused hardware push aims to cut the electricity cost of running Gemini models. The Information reports the new chip would generate many more tokens per unit of power, a metric that directly translates into lower operating expenses for Google Cloud customers. With the cloud segment posting $20 billion in Q1 2026 revenue and a 63 % year‑over‑year growth, the efficiency edge is critical to turning the massive $462 billion backlog into actual cash flow [3][2].
Google’s move mirrors a broader industry trend where AI firms build custom silicon to reduce reliance on Nvidia’s GPUs. OpenAI recently unveiled its Jalapeño inference chip, while Anthropic is courting Samsung for a new design [1]. Google’s claim of up to tenfold efficiency gains positions its TPU line as a stronger internal moat compared with Nvidia‑dominant alternatives, echoing CEO Sundar Pichai’s statement that “we own frontier models and own the silicon” [2]. The chip’s efficiency could also help Google compete with Amazon’s Trainium, which is marketed as 30 % cheaper per performance unit than comparable GPUs [2].
The rumor of Frozen v2’s efficiency helped lift Alphabet’s shares about 3 % on the morning after the report surfaced, reflecting investor optimism that the company can justify its $180‑$190 billion AI investment plan [1]. Analysts see the chip as a lever to mitigate the “compute‑constrained” outlook Pichai highlighted, especially as the bulk of TPU hardware revenue is expected to materialize in 2027 [2].
| Metric | Frozen v2 (projected) | Current TPU | Nvidia GPU |
|---|---|---|---|
| Tokens per power unit | 6‑10× higher | Baseline | Baseline |
The real test will be whether Frozen v2 can translate its projected efficiency into tangible cost savings for Google Cloud, thereby unlocking revenue from the $462 billion backlog and sustaining the company’s AI‑driven growth trajectory.
Coverage is mostly measured — 224 of 235 reports stay neutral.
Every Monday — the token unlocks, Fed dates & catalysts set to move crypto and markets this week. So you’re never blindsided.
Free · 3-min read · one-click unsubscribe
AI-assisted synthesis by the TrendWatcher Editorial Desk · sourced from 4 outlets · Jul 21, 2026 · How we report
Yes, after the policy change all backup data will count toward your Google Account storage limit, with an estimated average increase of about 40 MB per user.
Surging demand for AI computing power has outpaced existing infrastructure, leading to limited availability and the need for substantial capital expenditures on new data centers and custom chips.
Frozen v2 is targeted for deployment in 2028 as a specialized complement to existing TPUs, aiming to improve token‑per‑power efficiency by six to ten times.
It retains the 50 MP sensor array but adds AI‑based Pro Res Zoom for up to 100× digital zoom and a new 42 MP front camera with a 103° field of view and autofocus.
Google has committed to seven full years of Android and security updates, covering the device through 2032.