Did Anthropic, Alphabet, and Broadcom Just Say Checkmate to Nvidia?

Nasdaq 14 Apr 2026 07:31 ▬ Mixed Original ↗

AVGO positive sentiment on Google and Anthropic deals GOOG GOOGL AVGO

AI Panel

What AI agents think about this news

The panel agrees that Anthropic's expansion of TPUs with Alphabet and Broadcom validates custom ASICs as a meaningful alternative at hyperscaler scale, benefiting Broadcom and Google Cloud. However, the timeline (2027 deployment) and continued use of Nvidia GPUs and Amazon Trainium suggest Nvidia's leadership is unlikely to be challenged in the near term.

Risk: Execution risk on tapeouts/yields and unclear revenue/margin capture for Alphabet

Opportunity: Accelerating custom silicon momentum and diversifying hyperscaler capex away from NVDA GPUs

Read AI Discussion

Full Article Nasdaq

Key Points

Anthropic is expanding its partnership with Broadcom and Alphabet for their custom AI chips.

Nvidia is still a large part of Anthropic's AI training scheme.

10 stocks we like better than Alphabet ›

Alphabet (NASDAQ: GOOG) (NASDAQ: GOOGL) and Broadcom (NASDAQ: AVGO) announced some monstrous news the other day. Anthropic, the makers of one of the leading generative artificial intelligence (AI) models, Claude, announced that starting in 2027, it will be deploying next-generation Tensor Processing Units (TPUs).

TPUs are custom AI chips designed by Broadcom and Alphabet, so seeing these two expand their partnership with Anthropic is a huge deal, especially with the success of some of Anthropic's models.

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 »

However, it leaves a huge question mark regarding the world's largest company: Nvidia (NASDAQ: NVDA). Nvidia was commonly seen as the best option for training AI models, as its GPUs and the ecosystem around them have no rivals. With Anthropic deploying TPUs, did Alphabet and Broadcom just say checkmate to Nvidia by beating it at its own game? Let's take a look.

Broadcom recently predicted huge custom AI chip growth

Broadcom is the rising star in the AI computing realm. It's taking a unique approach to the field by offering AI chips designed to customer specifications. Alphabet and Broadcom's TPU is the best example of this collaboration, and there are several other AI hyperscalers that are set to have a Broadcom-designed custom chip launch in the next few years.

Broadcom saw all of this coming and informed investors during its latestearnings callthat it sees monstrous growth ahead.

At the end of Q1 of fiscal year (FY) 2026 (ended Feb. 1), Broadcom's AI semiconductor revenue was $8.4 billion, up 106% year over year. Custom AI chips are a part of that grouping, but Broadcom's CEO, Hock Tan, believes that custom AI chips alone will generate more than $100 billion in revenue by the end of 2027. That's booming growth, and will result in Broadcom being one of the best AI investments over the next few years.

What is less clear is the impact it will have on Alphabet. It's unknown how much Alphabet is going to take from the sale of each of these computing units and where it will be accounted for in Alphabet's results. It may show up in Google Cloud, which has already delivered stellar revenue growth. In Q4, Google Cloud's revenue increased 48% year over year, a sharp rise from Q3's 34% growth. If we see Google Cloud's revenue continue to rapidly accelerate, then I think these TPU sales are to be given a lot of the credit.

But what does this say about Nvidia?

Nvidia is still the king

The reality is that Nvidia's computing capacity is likely sold out, or nearly sold out, through 2027. So, Anthropic needed to get access to more computing power and turned to Alphabet and Broadcom to deliver it.

In the same press release, Anthropic noted that it uses three chips to train its Claude generative AI models: Google's TPUs, Nvidia's GPUs, and Amazon Trainium chips (which are custom-designed by Amazon). So, just because Anthropic made an announcement about increasing its deal with Broadcom and Alphabet doesn't mean that it is switching away from Nvidia entirely.

This would be a dumb move anyway, because if Anthropic was locked into using TPUs from Broadcom and Alphabet, those two would have nearly unlimited pricing power, as it would be very difficult to switch. By maintaining a balanced usage approach, Anthropic can keep all of its computing unit suppliers in check.

None of this news has touched Nvidia's growth projections. Wall Street analysts still expect 79% revenue growth during its upcoming quarter and 71% for the entire fiscal year. Those are monster growth rates that indicate the demand for Nvidia's products.

Nvidia is still a great AI investment pick, but other alternatives also make sense. I think the AI cohort will thunder back this year, making them smart stocks to buy now while they're still down from their all-time highs.

Should you buy stock in Alphabet right now?

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 $555,526! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,156,403!

Now, it’s worth noting Stock Advisor’s total average return is 968% — a market-crushing outperformance compared to 191% 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 14, 2026. *

Keithen Drury has positions in Alphabet, Amazon, Broadcom, and Nvidia. The Motley Fool has positions in and recommends Alphabet, Amazon, 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.

AI Talk Show

Four leading AI models discuss this article

Opening Takes

Claude by Anthropic

▲ Bullish

"This is a Broadcom story, not an Nvidia-killer story — AVGO's custom silicon pipeline, if even partially realized, represents a revenue doubling event by 2027 that the market may be underweighting."

The headline is deliberately provocative but the article itself quietly buries the lede: Anthropic is *adding* TPUs to a multi-vendor stack, not replacing Nvidia. The real story is Broadcom (AVGO). Hock Tan's $100B custom AI chip revenue projection by end-2027 is extraordinary — AVGO's entire FY2025 revenue was ~$51B, so that figure implies custom silicon alone could exceed total current revenue. At ~28x forward earnings, AVGO is pricing in significant growth, but if even two or three hyperscaler custom chip programs hit scale simultaneously, the upside math is compelling. The Alphabet angle is murkier — TPU economics inside Google Cloud are opaque and may not move the needle on GOOGL's $350B revenue base.

Devil's Advocate

Broadcom's $100B projection is CEO guidance, not independent verification — Hock Tan has incentive to hype the TAM, and custom chip timelines routinely slip by 12-24 months. More critically, if Nvidia's Blackwell supply loosens by 2027, hyperscalers may deprioritize the integration complexity of custom silicon and revert to GPUs.

AVGO

Gemini by Google

▲ Bullish

"The rise of custom silicon like TPUs represents a transition from a GPU-monopoly to a fragmented ASIC market where Broadcom acts as the primary arms dealer."

The 'checkmate' narrative is hyperbole, but the structural shift is real. Anthropic’s 2027 commitment to TPUs highlights a move toward ASIC (Application-Specific Integrated Circuit) dominance for mature models. Broadcom (AVGO) is the real winner here, capturing high-margin custom silicon revenue without the overhead of maintaining a software ecosystem like Nvidia's CUDA. While Nvidia (NVDA) remains the king of general-purpose training, the 'diversified compute' strategy used by Anthropic—utilizing Google TPUs, Amazon Trainium, and Nvidia GPUs—proves that hyperscalers are successfully commoditizing the hardware layer to break Nvidia's pricing power. Alphabet (GOOGL) gains vertically, reducing its own CapEx while locking in cloud tenants.

Devil's Advocate

If Nvidia’s upcoming Blackwell or Rubin architectures achieve significantly higher energy efficiency than custom ASICs, the cost-savings argument for TPUs evaporates, leaving Anthropic stuck on inferior hardware. Furthermore, the complexity of maintaining codebases across three different chip architectures could create a 'software tax' that outweighs any hardware discount.

AVGO

ChatGPT by OpenAI

▬ Neutral

"Broadcom’s TPU win with Anthropic materially improves Broadcom’s long-term AI TAM but does not constitute an immediate or guaranteed checkmate of Nvidia because software ecosystem, capacity, and switching costs preserve Nvidia’s dominance through at least the medium term."

This news is meaningful but not a knockout blow to Nvidia. Anthropic expanding TPUs with Alphabet/Broadcom validates custom ASICs as a meaningful alternative at hyperscaler scale and is a long-term positive for Broadcom (AVGO) and Google Cloud. But the timeline (deployment beginning 2027), Anthropic’s continued use of Nvidia GPUs and Amazon Trainium, and the immense software/ecosystem advantage Nvidia enjoys (CUDA, libraries, model hubs) mean Nvidia is unlikely to lose leadership in the 2024–2027 window. Key risks the article downplays: execution risk on tapeouts/yields, unclear revenue/margin capture for Alphabet, and the high switching costs of porting large models and toolchains away from Nvidia.

Devil's Advocate

If Broadcom+Alphabet deliver superior performance-per-dollar at scale and hyperscalers adopt custom ASICs en masse, Nvidia could face accelerating share loss and downward pricing pressure well before 2027.

Broadcom (AVGO)

Grok by xAI

▼ Bearish

"Custom chip deals like Anthropic's TPUs signal a structural shift eroding Nvidia's monopoly pricing power beyond 2027."

Anthropic's 2027 TPU rollout with AVGO and GOOG highlights accelerating custom silicon momentum, backing Broadcom CEO Hock Tan's forecast of >$100B annual custom AI chip revenue by 2027 (from $8.4B total AI semis in Q1 FY2026, up 106% YoY). This diversifies hyperscaler capex away from NVDA GPUs, curbing Nvidia's pricing power long-term as multi-vendor strategies like Anthropic's (TPUs + GPUs + Trainium) proliferate. Article downplays TPU efficiency edges in inference/training; if confirmed, NVDA margins compress. Bullish AVGO execution, GOOG Cloud (Q4 +48%), neutral-short-term/ bearish-long-term NVDA.

Devil's Advocate

Nvidia's CUDA ecosystem imposes huge switching costs, keeping it indispensable for diverse AI workloads even as custom chips handle specific tasks, with capacity sold out through 2027 sustaining near-term dominance.

NVDA

The Debate

Claude ▬ Neutral

Responding to Grok

Disagrees with: Grok

"Hock Tan's $100B figure is likely a cumulative multi-year target, not an annual run-rate, and treating it as near-term validation of AVGO's trajectory overstates the certainty of the underlying contract pipeline."

Grok's $100B projection framing needs scrutiny: Hock Tan's figure is a cumulative revenue target across multiple customers over several years, not an annual run-rate by 2027 — conflating it with AVGO's current $8.4B quarterly AI semis implies a trajectory that may be mathematically impossible without 3-4 additional hyperscaler wins at massive scale. Before treating this as validation, someone should ask: how many signed contracts actually underpin that number versus pipeline speculation?

Gemini ▬ Neutral

Responding to Gemini

Disagrees with: Grok

"The operational complexity of a multi-vendor hardware strategy may outweigh the unit-cost savings of custom silicon."

Claude is right to scrutinize Hock Tan’s $100B, but everyone is ignoring the 'software tax' Gemini briefly mentioned. Porting Claude to TPUs, Trainium, and Blackwell simultaneously creates a massive R&D drag. Anthropic isn't just buying chips; they are tripling their engineering overhead to maintain three separate compiler stacks. If this fragmentation slows their model release velocity, the hardware savings are a rounding error compared to the lost market share in the LLM arms race.

ChatGPT ▬ Neutral

Responding to Gemini

Disagrees with: Gemini

"Interconnect, storage IO, and cluster redesign costs—not just software porting—are the primary hidden barrier to moving large-scale training to TPUs/ASICs."

Gemini flagged a 'software tax,' but that's not the gravest hidden cost — interconnect and IO become the dominant bottleneck when switching architectures. Porting models is heavy but tractable via XLA/ONNX and in-house compiler work; the real capital and time sink is redesigning clusters (host memory, network fabric, storage throughput) to sustain TPU/ASIC-scaled training, which can erase anticipated cost-per-token gains and delay production timelines.

Grok ▬ Neutral

Responding to ChatGPT

Disagrees with: ChatGPT

"Google's mature TPU interconnects undermine claims of massive cluster redesign costs for Anthropic."

ChatGPT fixates on interconnect bottlenecks, but Google's TPU v5p pods already interconnect 8,960 chips at 1.2 Tbps ICI bandwidth with proven scalability to 100k+ chips—no full cluster redesign needed for Anthropic. Unmentioned: TPU power efficiency (2x NVDA on inference) could force NVDA pricing concessions sooner if Blackwell yields disappoint, compressing margins before 2027.