NVIDIA Data Center Revenue Trajectory, Memory Shortage Headwinds, and Marvell's Custom Silicon and Optical Surge

NVIDIA Data Center Revenue Trajectory, Memory Shortage Headwinds, and Marvell's Custom Silicon and Optical Surge

The artificial intelligence hardware cycle has reached unprecedented heights in the second quarter of fiscal 2027 (ended July 26, 2026), as NVIDIA Corporation (NVDA) and Marvell Technology, Inc. (MRVL) reported blowout results that shattered Wall Street estimates. NVIDIA delivered a historic, first-ever year-ahead forecast projecting 70% revenue growth for Fiscal 2028, while Marvell disclosed a massive, long-term custom silicon and warrant agreement with Google that is projected to yield up to $120 billion in cumulative revenues over six years at full performance.

However, both semiconductor giants warned that severe supply chain constraints—particularly surging memory component costs (specifically HBM) and packaging bottlenecks—remain the primary limiting factors on how quickly they can scale.

NVIDIA Q2 FY2027 Blowout and Historic Year-Ahead Guidance

NVIDIA reported record-breaking quarterly revenue of $96.22 billion, representing 106% year-over-year growth and comfortably beating Wall Street consensus of $92.2 billion. The engine remains the Data Center segment, which surged 117% year-over-year to $89.0 billion (exceeding estimates of $85.8 billion). Within data center revenue, Hyperscale cloud providers accounted for $48.7 billion (up 102% YoY), while Enterprise, Consumer Internet, and Sovereign AI (ACIE) surged 138% YoY to $40.3 billion.

In a rare move, NVIDIA provided its first-ever year-ahead forecast, projecting that revenue in Fiscal 2028 (ending January 2028) will grow by approximately 70%. This implies FY2028 revenue in the range of $690 billion to $700 billion, vastly exceeding the $570 billion (44% growth) that analysts had been modeling. CEO Jensen Huang emphasized that demand supports even greater numbers, but supply limits what they can confidently deliver:

"Even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%... We wanted to make sure that everybody has the same set of information. We've got a huge year coming up next year, and it's going to be pretty extraordinary."

Surging Memory Procurement and Margin Compression

NVIDIA's massive scale-up has triggered a sequential surge in supply and capacity commitments, which more than doubled from $119 billion in the prior quarter to $279 billion in Q2, primarily driven by long-term procurement of high-bandwidth memory (HBM). Surging memory prices are compressing gross margins: Q2 GAAP and non-GAAP gross margins came in at 75.0%, and NVIDIA guided Q3 gross margins down to 74.0%. CFO Colette Kress indicated that margins will bottom in Q4 at 71% to 72% before settling around 72% to 73% in FY2028. To offset these inflationary pressures, NVIDIA has begun raising hardware prices to customers by approximately 15%.

Furthermore, persistent HBM shortages have forced a major specification downgrade for NVIDIA's next-generation Rubin Ultra GPU. Initially previewed with up to 1TB of HBM, the Rubin Ultra is now slated to ship with just 192GB of HBM4 (8-hi)—which is actually lower than the regular Rubin GPU's 288GB of HBM4 (12-hi)—due to tight DRAM supply expected to persist through 2027.

Marvell Technology Q2 2027: Google Custom Silicon Deal and Optical Ramps

Marvell Technology (MRVL) also delivered record fiscal Q2 2027 results (ended August 1, 2026), posting net revenue of $2.739 billion (+37% YoY) and beating the midpoint of guidance. The record was entirely powered by the Data Center segment, which grew 46% year-over-year to $2.171 billion and now represents 79% of total company revenue (up from 74% a year ago).

Following these results, Marvell raised its full-year revenue outlook:

• FY2027 Revenue: Raised to ~$12 billion (approx. 45% growth), up from prior guidance of $11.5 billion.
• FY2028 Revenue: Raised to ~$18 billion (approx. 50% growth), up from prior guidance of $16.5 billion.
• Data Center Revenue: Now expected to grow ~60% in FY2027 and more than 60% in FY2028.

The Landmark Google TPU and Warrant Agreement

A key driver of Marvell's long-term trajectory is an expanded, long-term custom silicon agreement with Google spanning through fiscal 2033. Under this deal, Google received a warrant to acquire up to 7% of Marvell shares, tied to revenue milestones. The partnership spans a broad range of custom silicon programs attaching to the TPU ecosystem, including AI inference accelerators, storage controllers, network interface controllers (NICs), memory interface controllers, and near-memory compute.

Analysts project that the commercial agreement represents up to $120 billion in cumulative revenues over six years if all milestones are hit, averaging $18.5 billion per year at peak performance. While some initial revenue is reflected in FY2028, CEO Matt Murphy indicated that the massive step-ups will occur in FY2029 and beyond:

"This opportunity with this customer greatly, greatly increases the revenue opportunity for us in custom with some of that potentially starting in 2029... whatever you have modeled previously prior to the warrant for custom numbers definitely goes higher."

Optical Scale-Up and Celestial AI Ramps

Beyond custom ASICs, Marvell is seeing a massive acceleration in its scale-up optics portfolio, including Near-Package Optics (NPO) and Co-packaged Optics (CPO) utilizing silicon photonics.¹ This includes the rapid integration of Celestial AI’s CPO Photonic Fabric technology (projected to yield $150 million in FY2028). The aggregate scale-up optics and switching portfolio (including Marvell's UALink and ESUN switches) represents a multi-billion dollar lifetime revenue opportunity as hyperscalers build larger AI clusters that span multiple data centers.