Nvidia (NASDAQ: NVDA) is a $5tn company again, and this time Google’s TPUs are watching

Nvidia Corporation (NASDAQ: NVDA) closed at an all-time record of $208.27 on Friday, gaining 4.32 percent and pushing its market capitalisation back above $5 trillion for the first time since late October 2025. The move re-establishes Nvidia as the world’s most valuable publicly traded company, sitting roughly $1 trillion ahead of Alphabet Inc. The rally was triggered less by Nvidia-specific news and more by a blowout earnings print from Intel Corporation, which surged 23.64 percent in its best single-day performance since 1987 and reignited investor appetite across the entire semiconductor complex. The timing matters: the close lands days before the four hyperscalers, Microsoft Corporation, Meta Platforms Inc., Alphabet Inc., and Amazon.com Inc., report quarterly results that will set the next direction for AI capital expenditure expectations and, by extension, for Nvidia itself.

The headline number sits inside a larger structural story. Nvidia first crossed $5 trillion on October 29, 2025, then drifted lower as oil price spikes from the Iran conflict and broader large-cap technology de-risking pulled the stock off its highs. The April recovery has been swift. Nvidia is up roughly 20 percent over the past month and 95.68 percent over the past 12 months, with year-to-date gains of 10.28 percent. The Philadelphia Semiconductor Index, the SOX, has now logged an 18-session winning streak, with the broader rally adding tens of billions of dollars in value to peers including Broadcom Inc., Taiwan Semiconductor Manufacturing Company, Micron Technology, Advanced Micro Devices, Texas Instruments, and Intel.

What does Nvidia’s second crossing of the $5 trillion market cap signal about AI infrastructure spending in 2026?

The first $5 trillion print in October 2025 was a coronation moment. The second one, six months later, is a verdict on whether the AI capital cycle is durable. Investors are reading Friday’s close as an early bet that hyperscaler capex commitments are not peaking. Nvidia’s own fiscal Q4 2026 results in late February confirmed that hyperscalers accounted for just over 50 percent of data centre revenue, and combined hyperscaler capex for calendar 2026 is tracking towards $600 billion to $700 billion, with roughly three-quarters of that earmarked for AI infrastructure. Nvidia’s data centre segment delivered more than 91 percent of total quarterly revenue in the most recent print, and the Q1 FY27 revenue guide of $78 billion implied that demand was outrunning even the most bullish sell-side models.

The question for the upcoming hyperscaler earnings cycle is whether Microsoft, Meta, Alphabet, and Amazon raise, hold, or trim their multi-year capex envelopes. A raise validates Friday’s close. A hold is neutral but extends the runway. A cut, even a marginal one, would re-test the AI destabilisation fears that briefly shook Nvidia stock in February despite a record print. Sector strategists have noted that a quarter of the S&P 500’s year-to-date gains are attributable to Nvidia alone, which means any hyperscaler-driven repricing of AI capex risk would not stay contained to a single ticker.

How does the Intel earnings surprise change the competitive map for Nvidia and AMD in AI compute?

Intel’s quarterly earnings of 29 cents per share against consensus expectations of just one cent, paired with revenue of $13.58 billion versus the $12.42 billion consensus, was the immediate catalyst for Friday’s chip-wide rally. The ripple was disproportionate. Advanced Micro Devices closed up 14 percent and Qualcomm Inc. gained 11 percent, with Nvidia’s 4.3 percent advance carrying the largest absolute dollar weight given its market cap. The cleaner read, however, is that capital is rotating back into semiconductors as a thematic basket rather than picking individual winners and losers.

That basket framing helps Nvidia in the short term but introduces a longer-term complication. If Intel can stabilise its foundry economics and re-enter the AI compute conversation, and if AMD’s MI350 and MI400 series continue to win inference share, the assumption that Nvidia retains 90 to 92 percent of the AI accelerator market becomes harder to extend through 2027 and beyond. Wall Street is currently underwriting Nvidia at a consensus Strong Buy rating with a median price target of $264.11, which embeds a continuation of Blackwell-to-Rubin transition economics. That underwriting tightens if competitive share moves even five percentage points.

Why are Google TPUs and custom hyperscaler silicon a structural headwind that Nvidia must price into its forward growth case?

The competitive threat that did not exist in the same form during Nvidia’s October 2025 ascent is now firmly on the field. On April 22, Alphabet’s Google Cloud unveiled its eighth-generation Tensor Processing Units, splitting the architecture into a TPU 8t for training and a TPU 8i for inference. Google claims the new TPUs deliver 3x faster training, 2.8x improved price-to-performance, and up to 80 percent better performance per dollar versus the prior generation, with the ability to cluster more than one million chips together.

Two parallel data points sharpen the threat. Anthropic has committed to multi-gigawatt TPU capacity from Google, including a separate Broadcom Inc. arrangement giving Anthropic access to roughly 3.5 gigawatts of TPU-based compute starting in 2027. OpenAI has also been added as a TPU customer, even as it remains a major Nvidia buyer. Amazon Web Services continues to scale Trainium2, and Broadcom has signed a multi-year custom accelerator partnership with OpenAI for 10 gigawatts of capacity. Custom silicon is no longer a theoretical hedge against Nvidia pricing power. It is a deployed, in-production alternative with named anchor customers, and Broadcom is projected to control roughly 60 percent of the AI server custom ASIC market by 2027.

Nvidia’s counter-positioning rests on the Rubin architecture, deeper integration with hyperscaler networking through NVLink and Spectrum-X, and a software moat around CUDA that has so far proven sticky. Google itself confirmed it will offer Nvidia’s Vera Rubin systems later this year, signalling that TPU expansion is additive rather than purely substitutive. The strategic risk is not displacement. It is margin compression. Every workload that runs on TPUs, Trainium, or Broadcom-designed silicon is a workload Nvidia does not capture at premium gross margins.

What execution and regulatory risks could derail the Nvidia bull case from current levels?

Nvidia faces three converging risk vectors that the current price does not fully discount. The first is China exposure, where export controls continue to evolve and where the H20 product cycle has already been disrupted. The second is the law of large numbers. Maintaining 60 percent-plus revenue growth from a base of $215.9 billion in fiscal 2026 revenue is mathematically demanding, and any deceleration in the Rubin ramp could reset the multiple. The third is concentration risk. With over half of data centre revenue tied to four hyperscaler customers, each of whom is simultaneously building competing silicon, the bargaining dynamic is shifting in subtle ways that earnings calls do not yet fully reflect.

The market reaction layer adds nuance. Nvidia’s $208.27 close sits just below the all-time intraday high of $212.19 set on October 29, 2025. The fact that the stock is once again knocking against that ceiling, despite six months of competitive escalation and despite a brief AI bubble panic in February, is itself a statement. It says that capital is treating the structural AI compute build-out as a multi-year infrastructure cycle rather than a cyclical chip upturn.

Key takeaways on what the Nvidia $5 trillion close means for the company, its competitors, and the AI infrastructure cycle

• Nvidia’s record close at $208.27 reclaims the $5 trillion market cap threshold and restores its $1 trillion lead over Alphabet, but the catalyst was Intel’s earnings beat rather than Nvidia-specific news, signalling sector-wide capital rotation.
• The next directional read is the upcoming hyperscaler earnings cycle from Microsoft, Meta, Alphabet, and Amazon, where any change in 2026 AI capex guidance will move Nvidia more than any product announcement.
• Combined hyperscaler AI capex is tracking towards $600 billion to $700 billion in calendar 2026, with hyperscalers accounting for over 50 percent of Nvidia data centre revenue.
• The Philadelphia Semiconductor Index’s 18-session winning streak indicates that capital is treating chips as a thematic basket, which lifts Nvidia in the short term but blurs competitive differentiation.
• Google’s eighth-generation TPU 8t and TPU 8i, launched April 22, plus Anthropic’s multi-gigawatt TPU commitment and OpenAI’s expanded TPU access, mark a step change in custom-silicon viability against Nvidia GPUs.
• Broadcom is on track to control roughly 60 percent of the AI server custom ASIC market by 2027, making it the structural beneficiary of any hyperscaler diversification away from Nvidia.
• Advanced Micro Devices has captured roughly 8 percent of the AI accelerator market with the MI350 and MI400 series, narrowing what was effectively a Nvidia monopoly twelve months ago.
• Wall Street’s median price target of $264.11 prices in successful Blackwell-to-Rubin transition economics, leaving limited room for execution slips, China-related disruption, or hyperscaler capex cuts.
• The risk for Nvidia is no longer displacement but margin compression, as every TPU, Trainium, and Broadcom-designed accelerator workload is revenue Nvidia does not capture at premium gross margins.
• A second visit to $5 trillion validates the AI infrastructure thesis, but durability now depends on whether Rubin can defend pricing power as custom silicon scales from 2026 into 2027.