Nvidia's Two Clocks: Supply Chain Ramps Over Years While Goldman Sees 300

Goldman Sachs reaffirmed its buy rating on Nvidia and kept a 300-dollar price target, but analyst James Schneider expects the latest announcements to do little for the share price in the near term. The distinction matters: Goldman is separating the operating story from any immediate catalyst, and the stock's 2.3% gain today appears to lean on confirmed buy calls and demand signals rather than on any single headline.

The linchpin of Goldman's case is customer-side economics. Nvidia's customers, in the firm's view, are earning strong unit economics — a point that carries more weight for investors than a simple confirmation of large capital budgets. Whether installed compute is used profitably determines how sustainable further spending will be. Goldman's read is positive, though it draws no short-term trading impulse from it.

Buybacks and a Business Line That Outgrew Last Year

As a supporting argument, the firm points to roughly 235 billion dollars in authorized share buybacks. That is an approved framework, not a completed program — a distinction that matters for how the figure should be read. Growth beyond classic data center workloads adds another leg: Nvidia reported 11 billion dollars in Physical AI revenue for the first half of fiscal 2027, already above the full fiscal 2026 figure.

The operating numbers back the confidence. Second-quarter fiscal 2027 revenue came in at 96.2 billion dollars, up 106% year over year. Melius also reiterated its buy recommendation today, calling results from SpaceX and Super Micro a positive signal for Nvidia — external developments that reinforce the demand picture.

Six Months to Fourteen Days

On the supply side, Nvidia is pushing on several links of the chain at once. The company announced additional capacity with partners across substrates, advanced chip packaging and testing — a multi-stage effort rather than a single fab project. IBIDEN, Samsung Electro-Mechanics, Unimicron, Amkor, KYEC and SPIL are all involved, and substrate capacity is slated to rise 2.5-fold within three years.

Downstream, Amkor aims to expand advanced packaging capacity by more than 300% by the end of 2028, while KYEC is to build test facilities in the United States. Nvidia is backing that with a 1.1 billion dollar prepayment to reserve capacity — evidence it is willing to put money behind partner commitments, not just announce them.

A second lever targets deployment itself. The DSX POD, unveiled today, is a modular reference design for complete Vera-Rubin AI factories; partners will build, integrate and validate the systems before shipment. On-site setup could fall from six months to 14 days, provided the necessary power and infrastructure are already in place at the site. That makes it a conditional performance promise, not a blanket commitment for every data center project.

The Gartner Gap

Set against Goldman's optimism is Gartner's industry-wide view. The research house sees a substantial gap between the data center revenue operators will need and the revenue expected from AI applications: cumulatively, operators would have to generate 8.5 trillion dollars by the end of 2030 to justify their investments, against at most 3.6 trillion dollars in attributable revenue.

That forecast covers the sector, not Nvidia specifically, but it frames the risk sitting behind the investment boom. Both of today's threads carry the same caveat — the capacity buildout stretches over multiple years and the faster installation depends on prepared sites, so the announcements set targets and conditions rather than proving they have been met. Goldman's buy case rests on customer economics holding up; the long-term test is whether that economics survives an infrastructure base that keeps growing.