By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel discusses the potential of Anthropic's multi-GW TPU commitment with Broadcom and Google Cloud, but consensus is divided due to significant risks such as execution challenges, power constraints, and software porting friction.
Risk: Software porting friction and power constraints
Opportunity: Potential cost-efficient inference and strategic advantage
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 →
Alphabet and Broadcom's TPUs are starting to attract more clients.
Both Alphabet and Broadcom are seeing huge growth.
In the world of generative artificial intelligence (AI), few companies generate as much buzz as Anthropic. For example, its Claude platform is often the leading platform for assisting coders, and its latest model, Mythos, couldn't even be released to the general public due to its potential threat to cybersecurity. It's at the top of the food chain right now, and any company partnering with Anthropic is often seen as a leader. If their equipment is good enough for Anthropic, the thinking goes, it's likely about the best available.
Recently, Anthropic made an announcement regarding its usage of Tensor Processing Units (TPUs), which were created through a joint venture between Broadcom (NASDAQ: AVGO) and Alphabet (NASDAQ: GOOG) (NASDAQ: GOOGL). Both of these companies stand to benefit from increased TPU usage, and each looks like a phenomenal investment.
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
Anthropic announced that starting in 2027, it will be using multiple gigawatts of computing power of next-generation TPUs. Long-term commitments like these help give investors clarity for what to expect from companies in the years ahead.
This partnership, alongside several others that Broadcom has announced, will help it continue to deliver impressive revenue growth. By the end of 2027, Broadcom expects its custom AI chip business to generate more than $100 billion annually. Its AI semiconductor division (which includes other products outside of custom AI chips) generated $8.4 billion in revenue last quarter (up 106% year over year). There's a ton of growth ahead, and growing partnerships with AI leaders like Anthropic are a good sign for the future.
Alphabet is recognizing TPU revenue through its Google Cloud division, and this segment is catching fire. Last quarter, its revenue rose 48% year over year. That's rapid growth for a legacy company, and expanded TPU partnerships with Anthropic and others will lead to continued strong growth for this important segment within Alphabet.
Both of these stocks are fantastic investment options in the AI space. Instead of being in the leadership position like Nvidia, they are comfortably in the challenger role and only looking to take market share, which they appear to be doing. Anthropic will continue using Nvidia hardware as well, but Nvidia no longer has this massive AI growth segment to itself.
Still, I think a well-balanced approach of owning several of these companies is the best call for most AI investors.
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 $511,411! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,238,736!
Now, it’s worth noting Stock Advisor’s total average return is 986% — a market-crushing outperformance compared to 199% 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 April 22, 2026. *
Keithen Drury has positions in Alphabet, Broadcom, and Nvidia. The Motley Fool has positions in and recommends Alphabet, Broadcom, 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
"The shift toward custom silicon ASICs is the most significant threat to Nvidia's current margin profile, as it commoditizes the underlying hardware layer."
The Anthropic-Google-Broadcom partnership is a structural shift, not just a headline. By committing gigawatts of capacity to custom TPUs by 2027, Anthropic is signaling a strategic pivot away from total Nvidia dependency to optimize for cost-to-compute ratios. For Broadcom (AVGO), this validates their ASIC (Application-Specific Integrated Circuit) model as the primary alternative to general-purpose GPUs. Alphabet (GOOGL) gains a massive 'dogfooding' win, proving Google Cloud is a viable ecosystem for frontier model training. However, the market is pricing this as pure incremental upside, ignoring the massive CapEx burden and the risk of 'model collapse' if custom silicon fails to keep pace with rapid architectural changes in transformer models.
If model architectures evolve away from current TPU-optimized designs, these long-term gigawatt commitments could become 'stranded assets' that drag down margins rather than driving growth.
"Anthropic's TPU pledge positions Broadcom's custom AI chip business for $100B+ annual revenue by 2027, accelerating high-margin growth beyond current $8.4B quarterly run-rate."
Anthropic's multi-GW next-gen TPU commitment from 2027 validates the Alphabet-Broadcom platform as a Nvidia alternative, with Broadcom's custom AI chips eyeing $100B+ annual revenue by then (from $8.4B Q AI semis, +106% YoY). This high-margin segment could re-rate AVGO (35x forward P/E) higher on confirmed hyperscaler diversification. Alphabet's Google Cloud (+48% YoY to $12B) gains TPU tailwinds, but trails AWS ($29B) and Azure. Missing context: Anthropic still leans on Nvidia GPUs; TPU shift may prioritize cost-efficient inference over training.
The 2027+ timeline offers zero near-term revenue lift amid AI capex scrutiny (e.g., MSFT's recent moderation), and Broadcom's $100B projection hinges on unproven multi-year execution against Nvidia's 90%+ GPU dominance.
"A 2027 commitment is optionality, not revenue—and TPU competitiveness vs. Nvidia's ecosystem moat remains unproven at scale."
The article conflates announcement with execution. Anthropic committing to 'multiple gigawatts' of next-gen TPUs *starting 2027* is a 3-year forward commitment—not revenue today. Broadcom's $100B custom AI chip target by 2027 is aspirational guidance, not booked business. The real risk: TPU economics must remain competitive vs. Nvidia's entrenched software stack and manufacturing scale. Google Cloud's 48% YoY growth is real, but TPU attach rates within that segment aren't disclosed—we don't know if TPUs are driving growth or if it's other services. The article also omits that Anthropic remains heavily dependent on Nvidia chips today and may diversify suppliers rather than consolidate.
If Broadcom and Google can't achieve cost parity or superior performance per watt vs. Nvidia's next-gen H200/Blackwell, this 2027 commitment becomes a sunk cost write-down, and the article's 'challenger taking share' narrative collapses into margin compression.
"Anthropic’s planned long-run TPU uptake could be a durable tailwind for Alphabet and Broadcom, but the thesis hinges on sustained Anthropic demand and continued AI hardware pricing power in a Nvidia-led ecosystem."
The article frames Anthropic’s TPU partnership as a clean growth driver for Alphabet and Broadcom, but the reality is more nuanced. AI hardware demand is volatile, capex-intensive, and highly dependent on Anthropic’s adoption curve and broader cloud demand. A multi-gigawatt commitment for 2027 sounds large, yet it hinges on sustained off-take, favorable utilization, and margin recovery for Broadcom’s AI chips, against Nvidia’s entrenched lead. Alphabet’s cloud growth is real, but TPU revenue mix and profitability aren’t disclosed. Regulatory, supply-chain, and AI-safety hurdles could delay deployments or temper spend. If the demand path proves durable, the payoff could be meaningful; otherwise, gains may be limited.
The strongest counter: even with Anthropic on TPUs, the incremental revenue is uncertain and may be smaller than advertised; Nvidia's ecosystem lock-in and potential shift to in-house silicon could cap Broadcom/Alphabet upside.
"The software migration cost from CUDA to TPU-native stacks is the primary, overlooked barrier to Anthropic's successful transition."
Claude is right to flag execution risk, but everyone is ignoring the 'software moat' factor. Nvidia’s CUDA isn't just hardware; it’s a developer ecosystem that TPUs struggle to replicate. Even if Broadcom delivers the silicon, Anthropic’s engineers face massive friction porting models to Google’s proprietary stack. This isn't just a capex play—it's a massive, multi-year R&D tax on Anthropic. If the developer migration fails, those gigawatts will sit idle, regardless of the hardware's performance-per-watt.
"Gigawatt-scale TPU commitments exacerbate unresolved data center power constraints, risking stranded capex before software issues materialize."
Gemini rightly flags CUDA moat friction, but everyone's missing the power crunch: multi-GW TPUs from 2027 amplify data center energy bottlenecks (US grids strained at 1-2GW per site). Alphabet's capex balloons amid DOE permitting delays; Broadcom's ASICs won't matter if facilities can't power up. This strands assets faster than software porting—AVGO/GOOGL re-rating hinges on utility-scale nuclear ramps nobody mentions.
"Power logistics matter, but execution speed and training velocity gains matter more—neither is disclosed."
Grok's power constraint is real, but underspecified. A 2027 multi-GW TPU deployment doesn't mean simultaneous activation—Anthropic likely staggers rollout across regions/quarters. More critical: neither Grok nor Gemini quantified the actual porting friction. CUDA dominance matters, but Anthropic's scale (Claude inference) may justify custom optimization. The real tell: does Anthropic's training velocity *accelerate* on TPUs vs. Nvidia, or does it stall? That determines whether this is strategic or a $billions sinkhole.
"Utilization risk and grid/cooling costs will determine the economics of this 2027 mega-TPU rollout, not the headline capacity."
Your energy risk framing is real but incomplete—the bigger near-term issue is utilization risk and grid/cooling economics that could compress margins even with multi-GW capacity. If Anthropic’s uptake lags or regional power constraints throttle ramp, the planned 2027 capacity becomes idle or requires heavy capital recycling. In addition, regulatory and tariffs on grid upgrades could throttle cost advantages, undermining the thesis that AVGO/GOOGL gain a durable margin uplift.
The panel discusses the potential of Anthropic's multi-GW TPU commitment with Broadcom and Google Cloud, but consensus is divided due to significant risks such as execution challenges, power constraints, and software porting friction.
Potential cost-efficient inference and strategic advantage
Software porting friction and power constraints