Alphabet Inc. (GOOGL) (GOOG) reports second-quarter earnings this week, and Wall Street will be watching AI spending, cloud growth, and advertising. But there's another question that might matter just as much: Can Google make artificial intelligence cheaper to run?
According to a report from The Information, Alphabet is developing a new server chip, internally called Frozen v2, designed specifically to run Gemini models more efficiently. If the report is accurate, this chip could represent Google's next move in the AI race—not by building larger models, but by lowering the cost of serving them. That could make Alphabet's earnings call about far more than just revenue and earnings per share.
The AI Race Is Shifting To Efficiency
Over the past two years, investors have largely measured AI leadership by one metric: spending. Alphabet, Microsoft Corp. (MSFT), Meta Platforms Inc. (META), and Amazon.com Inc. (AMZN) have collectively committed hundreds of billions of dollars toward AI infrastructure, while NVIDIA Corp. (NVDA) has emerged as one of the biggest beneficiaries by supplying the GPUs powering that buildout.
Frozen v2 hints at the next phase. According to the report, Google engineers believe the new chip could deliver between six and 10 times more tokens per unit of power than the company's latest Tensor Processing Units (TPUs) by embedding portions of Gemini's architecture directly into the silicon. If those projections hold, the implications extend well beyond Google's chip business.
Running AI models—known as inference—is quickly becoming one of the largest operating expenses for hyperscalers as millions of users generate prompts every day. Improving efficiency could allow Google to serve more AI requests while consuming less power and fewer computing resources.
Why Investors Should Listen Closely
That makes Alphabet's earnings call an opportunity for management to discuss more than just AI investments. Investors will be listening for any commentary on custom silicon, inference workloads, TPU deployment, and capital spending. Even modest updates could provide clues about how Google plans to manage AI costs as Gemini adoption expands.
The report also comes as investors increasingly question whether hyperscalers can continue raising AI capital expenditures indefinitely. If Google can improve AI economics through better hardware rather than simply buying more compute, it could reshape how Wall Street evaluates future AI spending.
The Bigger Story Isn't Another AI Chip
Alphabet has built custom AI chips for years, making Frozen v2 less significant as a product announcement than as a strategic signal. The next competitive advantage in AI may not come from training ever-larger models. It may come from making those models dramatically cheaper to operate.
That's why this week's earnings call matters. Wall Street already knows Google is spending aggressively to compete in AI. What investors don't yet know is whether the company has found a way to generate more AI output without proportionally increasing its infrastructure costs. If Frozen v2 is part of that answer, Alphabet's earnings could reveal that the next battle in AI isn't just about building smarter models—it's about building more efficient ones.
