By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel has mixed views on Alphabet's TPU external sales strategy. While some see potential in the 'sovereign AI' market and believe Google can capture a significant portion of the AI infrastructure market, others argue that the addressable market is smaller and more fragmented than projected, and that Nvidia's CUDA moat and hyperscalers' vertical integration pose significant challenges.
Risk: Structural fragmentation of the market due to hyperscalers vertically integrating their own silicon, which could cap Alphabet's realistic total addressable market far below $100B by 2030.
Opportunity: The 'sovereign AI' market, where non-hyperscaler entities seek alternatives to Nvidia's hardware monopoly, presents a potential opportunity for Alphabet to provide a 'good enough' hardware alternative and prevent vendor lock-in.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Nvidia (NASDAQ: NVDA) stock has advanced 1,300% since the artificial intelligence (AI) boom began in January 2023. The company's success, in terms of both financial results and share price appreciation, has been driven by its dominance in artificial intelligence accelerators, a market projected to top $300 billion this year.
Alphabet (NASDAQ: GOOGL) (NASDAQ: GOOG) has dabbled in AI accelerators for over a decade, but the company recently started selling custom silicon directly to customers, marking a more deliberate attempt to compete with Nvidia. Read on to learn more.
Where to invest $1,000 right now? Our analyst team just revealed what they believe are the 10 best stocks to buy right now, when you join Stock Advisor. See the stocks »
Nvidia invented the graphics processing unit (GPU) in 1999. Those chips, originally built to render realistic video game and computer graphics, have become the industry standard for accelerating complex data center workloads, such as artificial intelligence (AI), because they perform trillions of calculations per second.
Last year, Nvidia accounted for more than 80% of AI accelerator sales, according to Silicon Analysts. But its market share is likely to fall in the years ahead as more companies turn to custom silicon solutions. Three of its largest customers -- Amazon, Microsoft, and Alphabet -- have deployed chips purpose-built for AI.
Alphabet was the first to pursue a custom silicon strategy, and it remains the most significant threat to Nvidia. In 2016, Alphabet deployed its first tensor processing unit (TPU), a chip designed specifically for the matrix and vector-based math needed to build and run AI models.
Initially, Alphabet limited TPUs to internal use cases, but the company made its custom silicon accessible to Google Cloud customers in early 2018. In the second quarter, Alphabet began selling TPUs directly to clients for use in external data centers, positioning itself as a more direct threat to Nvidia.
Nvidia is unlikely to cede its dominance in the AI accelerator market for two reasons. First, the company has a significant competitive advantage in CUDA, a software platform comprising hundreds of code libraries and frameworks (building blocks) that help programmers write GPU-accelerated applications.
CUDA is the main reason Nvidia GPUs have become the dominant AI accelerators, and the proprietary nature of the platform constitutes a durable economic moat. "Once a team has built pipelines on CUDA, switching to another platform is prohibitively expensive," explains VentureBeat.
Second, TPUs run far fewer algorithms than GPUs because they are built for specific tasks. "While TPUs excel at specific deep learning workloads, they are far less flexible," according to VentureBeat. That means Nvidia GPUs could immediately run a new AI technology if one were invented tomorrow, but the same is not true of Alphabet TPUs.
Gil Luria, head of technology research at D.A. Davidson, sees custom silicon as a massive growth opportunity for Alphabet. Anthropic and Meta Platforms have agreed to spend billions of dollars on TPUs in the years ahead, and Alphabet recently announced a joint venture with Blackstone to build a TPU cloud business.
Looking ahead, Lura says Alphabet could eventually capture 20% of the AI infrastructure market, which would value its chips business somewhere around $900 billion. Meanwhile, Morgan Stanley analysts expect custom silicon (primarily Alphabet's TPUs) to account for 24% of AI accelerator sales in 2030, up from 15% today.
Those estimates suggest Nvidia will retain its dominance, but they also underscore the massive opportunity that sits before Alphabet. AI accelerator spending is projected to reach $600 billion in 2030, which means Alphabet could bring in more than $100 billion in revenue annually from TPU sales alone by the end of the decade.
I think investors are overlooking this opportunity. Alphabet trades at 19 times earnings, well below the three-year average of 25 times earnings. That valuation is cheap for a company whose earnings are projected to grow at 14% annually over the next three years. That's why Alphabet shares could soar as the company targets the AI accelerator market.
Before you buy stock in Alphabet, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Alphabet wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $395,463! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,268,290!
Now, it’s worth noting Stock Advisor’s total average return is 927% — a market-crushing outperformance compared to 211% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of August 5, 2026. *
Trevor Jennewine has positions in Nvidia. The Motley Fool has positions in and recommends Alphabet, Blackstone, Meta Platforms, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Alphabet's TPU push is credible but the article's $100B revenue scenario by 2030 dramatically overstates near-term displacement of Nvidia's CUDA ecosystem."
Alphabet's move to sell TPUs externally is real but the $100B+ revenue by 2030 projection assumes it captures 20% of a $600B market while overcoming Nvidia's CUDA moat. GOOGL at 19x forward earnings (vs 25x 3yr avg) looks optically cheap against 14% EPS growth, yet the article downplays that TPUs remain workload-specific and that hyperscalers (AMZN, MSFT) are also building their own silicon. Nvidia's 80%+ share isn't vanishing quickly. The bullish case requires Alphabet to scale external TPU sales far beyond current Google Cloud traction without margin dilution.
The strongest case against is that Alphabet has been "targeting" this market for a decade with limited external success; if TPUs truly threatened Nvidia, we'd have seen far more than two big named customers (Anthropic, Meta) and actual disclosed revenue. The $900B chip business valuation is pure extrapolation that ignores execution risk and potential price wars in custom silicon.
"The market is underestimating the potential for Alphabet to capture significant AI infrastructure share, which justifies a re-rating of its P/E multiple toward its historical 25x average."
Alphabet’s pivot to externalizing TPU sales is a strategic masterstroke, but the market is mispricing the execution risk. While the article highlights a 19x forward P/E, this discount reflects investor skepticism regarding Google’s ability to transition from a software-first cloud provider to a hardware-as-a-service vendor. Scaling custom silicon requires massive capital expenditure and a high-touch enterprise sales force that Google has historically lacked. If they capture even 10% of the AI infrastructure market, the multiple expansion to 25x is justified, but the 'moat' provided by CUDA makes this an uphill battle. Investors should look for margin compression in the near term as Google subsidizes TPU adoption to lure developers away from Nvidia's ecosystem.
Google’s TPU strategy may ultimately fail to gain traction outside of their own cloud, as developers prioritize the massive, mature software ecosystem of Nvidia’s CUDA over the marginal cost savings of custom silicon.
"The article mistakes a real but nascent product line (TPU external sales) for a $100B+ revenue stream, while ignoring that Alphabet's current valuation discount reflects legitimate concerns about search monetization and AI capex ROI, not hidden chip upside."
The article conflates two separate narratives: Alphabet's valuation (19x vs. 25x historical) and TPU upside. The valuation argument is weak—the 19x multiple likely reflects slowing search growth and AI capex drag, not a hidden discount. The TPU story is real but vastly overstated. Morgan Stanley's 24% custom silicon share by 2030 doesn't mean Alphabet captures it; the article admits TPUs are less flexible than GPUs. Anthropic and Meta committing to TPUs is noteworthy, but neither is a customer—they're building their own chips or using Nvidia. The $100B+ revenue claim by 2030 requires flawless execution, no competitive response, and demand that hasn't materialized yet. Meanwhile, Nvidia's CUDA moat is understated—it's not just switching cost, it's ecosystem lock-in that grows stronger as more AI engineers train on it.
If Alphabet executes flawlessly on TPU external sales and captures even 15% of a $600B market by 2030, that's $90B in new revenue at 40%+ margins—a genuine re-rating catalyst that the market hasn't priced in at 19x forward earnings.
"External TPU monetization is not guaranteed and depends on a rapid, scalable shift in AI infrastructure that may fail to materialize against Nvidia's entrenched ecosystem."
The article paints Alphabet as a future TPU powerhouse that could grab 20% of AI infrastructure and earn $100B+ in TPU revenue by 2030, with a 19x multiple vs 25x historic. Yet the real world requires external demand for TPUs at cloud scale, not just internal use; CUDA's software moat and Nvidia's ecosystem make broad adoption unlikely quickly. TPUs are task-specific; developers may not rearchitect pipelines for an arguably narrower advantage. Alphabet will need heavy capex, margin discipline, and cloud-scale sales to outpace Nvidia on price and performance. The market's huge TAM is uncertain, and the Blackstone JV and partner bets add execution risk; the stock's premium remains fragile.
External TPU monetization hinges on a rapid, large-scale shift in AI infrastructure that may never materialize given Nvidia's CUDA moat and the entrenched software ecosystem. The biggest risk is that the TAM is overestimated and Alphabet can't monetize TPUs fast enough to justify the stock's multiple.
"Vertical integration by other hyperscalers structurally shrinks Alphabet's external TPU TAM below consensus projections."
Claude's concession on flawless execution yielding $90B revenue at 40%+ margins ignores the capex intensity required to scale TPU fabs and global availability. Hyperscalers like AMZN and MSFT are vertically integrating their own silicon (Trainium, Maia), shrinking the addressable external market below Morgan Stanley's 24% custom share. This isn't just execution risk—it's structural fragmentation that caps Alphabet's realistic TAM far below $100B by 2030.
"The demand for non-Nvidia hardware from non-hyperscalers creates a floor for TPU adoption that the current focus on CUDA-only developer preferences ignores."
Grok, you're missing the 'sovereign AI' angle. Nations and large enterprises are terrified of Nvidia's supply chain bottleneck and pricing power. This isn't just about developers choosing CUDA; it's about a desperate, non-hyperscaler market looking for an alternative to Nvidia's hardware monopoly. Alphabet doesn't need to beat Nvidia's ecosystem; they just need to provide a 'good enough' hardware alternative that prevents vendor lock-in. That demand is being severely underestimated by everyone here.
"Sovereign demand for TPU alternatives is real but too fragmented and price-sensitive to offset the TAM loss from hyperscaler vertical integration."
Gemini's 'sovereign AI' angle has real teeth, but it conflates two markets. Non-hyperscalers avoiding Nvidia lock-in ≠ adopting Alphabet TPUs at scale. Nations want alternatives; they don't automatically want Google's hardware or cloud dependency. The addressable market for 'good enough' alternatives exists, but it's fragmented, price-sensitive, and geopolitically volatile—not the $100B+ runway the article implies. Grok's structural fragmentation point stands: hyperscalers self-integrating shrinks external TAM faster than sovereign demand grows it.
"Sovereign AI demand is too slow and fragmented to unlock a large external TPU TAM, and margin pressure from subsidies would sour Alphabet's upside."
Gemini, the sovereign AI angle adds flavor but isn't a demand lever that scales fast. Governments buy through complex procurement, not just chips, and their demand cycles are longer, with security/compliance overhead. Alphabet would need a credible, high-service ecosystem to deter lock-in, matching Nvidia's software moat—unlikely quickly. If hyperscalers push even deeper internal silicon, external TPU TAM could be far smaller, magnifying margin compression from subsidies.
The panel has mixed views on Alphabet's TPU external sales strategy. While some see potential in the 'sovereign AI' market and believe Google can capture a significant portion of the AI infrastructure market, others argue that the addressable market is smaller and more fragmented than projected, and that Nvidia's CUDA moat and hyperscalers' vertical integration pose significant challenges.
The 'sovereign AI' market, where non-hyperscaler entities seek alternatives to Nvidia's hardware monopoly, presents a potential opportunity for Alphabet to provide a 'good enough' hardware alternative and prevent vendor lock-in.
Structural fragmentation of the market due to hyperscalers vertically integrating their own silicon, which could cap Alphabet's realistic total addressable market far below $100B by 2030.