Reports indicate NVIDIA customers face a 15% price increase on accelerators, while Google allocates $200 billion to capture market share with custom silicon.
Some of Nvidia’s (NASDAQ:NVDA | NVDA Price Prediction) largest customers have been warned that AI servers built around Grace Blackwell and Vera Rubin chips could cost more than 15% extra in early 2027 as memory prices climb. Alphabet (NASDAQ:GOOG, NASDAQ:GOOGL) is among the data center operators exposed to those increases, placing it in a unique strategic position.
Alphabet purchases Nvidia hardware to serve enterprise customers who require it, while simultaneously selling a competing infrastructure stack built around its own tensor processing units. Management has committed $195 billion to $205 billion in capital spending this year on AI infrastructure, and a new partnership with Marvell (NASDAQ:MRVL) could generate up to $120 billion in custom-chip sales through fiscal 2033. Two names stand out as the clearest beneficiaries: Alphabet and Marvell Technology.
Google Cloud revenue grew 82% in the June quarter to $24.8 billion, and cloud backlog reached $514 billion. Alphabet expects to recognize just over 50% of that backlog as revenue over the next 24 months. CEO Sundar Pichai has framed the company’s offering as accelerator-agnostic, stating that Alphabet provides “the industry’s broadest range of accelerators from Google and NVIDIA, including the latest NVIDIA platforms and our TPUs, which deliver strong price performance.” This posture is strategically vital because it allows Alphabet to absorb Nvidia price increases while monetizing its alternatives simultaneously.
The capital program weighs heavily on short-term financials. Second-quarter free cash flow was negative $5.9 billion, long-term debt rose to $98.2 billion, and Alphabet raised roughly $70 billion in combined equity and debt to fund the buildout. Despite these near-term pressures, investors have rewarded the strategy. Shares of Alphabet are up 69.37% over the past year and closed at $348.06 on August 24. The market treats the TPU stack as a genuine second revenue engine rather than merely a defensive hedge.
On the supplier side, Marvell reported first-quarter fiscal 2027 revenue of $2.42 billion, up 27.57% year over year, with data center operations accounting for 76% of the total. Second-quarter guidance implies roughly 35% annual growth. CEO Matt Murphy told investors that Marvell has “custom engagements across the board at all the U.S. hyperscalers,” and noted that the custom business alone has a path to “over $10 billion in revenue in fiscal 2029.” These projections are grounded in designs already won, rather than speculative pipeline estimates.
Marvell has also expanded its relationship with Nvidia around silicon photonics and NVLink Fusion. Murphy explained that the technology “allows Marvell to build custom chips and networking semiconductors that can seamlessly interface with NVIDIA infrastructure.” Crucially, the same optics and interconnect products are sold into both TPU and Blackwell racks.
A reported 15% surcharge on Nvidia servers is likely insufficient on its own to trigger mass migration, but it significantly shifts negotiating leverage. Every enterprise renewing a multi-year cloud contract now has a credible reason to price out TPU capacity alongside GPU capacity. Furthermore, Alphabet does not need outside customers to justify the massive buildout. Between Search, YouTube, Gemini training runs, and frontier research, Google can absorb most of what it is constructing even if external TPU demand grows more slowly than the $514 billion backlog implies. Outside adoption represents pure upside on top of that baseline.
For investors seeking AI infrastructure exposure without having to pick between architectures, Marvell presents a cleaner bet. Whether hyperscalers lean harder on Nvidia GPUs or on their own XPUs, Marvell supplies the optics, interconnect, and often the custom silicon itself. (We profiled seven suppliers doing this kind of work, from power to networking, in a free report you can grab here.) Murphy stated the company wants to be “one of the big winners in this AI cycle.”
Alphabet is ultimately better positioned for this moment. The reported Nvidia price move gives Google’s sales team something concrete to point to when pitching TPU capacity, and the Marvell deal locks in supply for years, even if internal demand alone would not fill the fabs. Marvell benefits regardless, which is why its stock has already run harder.
Contact [email protected] for any questions or corrections.