Jensen Huang walked off the GTC stage on Monday having simply projected at the very least $1 trillion in chip income by 2027. Wall Road analysts spent Tuesday calling it a flooring, not a ceiling. And Nvidia (NVDA) inventory sat there, barely transferring, buying and selling proper the place it was earlier than the entire thing began.
That disconnect tells you one thing necessary about the place the Nvidia story stands proper now.
The bull case just isn’t in dispute. What analysts at the moment are zeroing in on is the subsequent battle Nvidia has to win, one it has by no means needed to battle earlier than. Coaching AI fashions was the primary struggle. Inference, operating these fashions at scale in the actual world, is the subsequent one. And in contrast to coaching, inference has actual competitors.
Why inference modifications every little thing
For 3 years, Nvidia owned AI as a result of it owned coaching. Its market share in AI coaching chips sits at roughly 90%, and that place stays largely intact heading into 2026.
However the trade is shifting quick. As AI strikes from the lab into manufacturing, the workload that issues most is altering. Inference is the place the majority of future compute demand will land. And it has a basically completely different economics profile than coaching.
Inference requires decrease peak efficiency however a lot greater quantity, and it operates on tighter margins. That makes it a neater goal for cheaper options. AMD’s MI400, Amazon’s Trainium3, and Google’s TPU are all positioning themselves as lower-cost choices for inference workloads, aiming at hyperscalers underneath intense stress to chop the associated fee per AI token.
What Nvidia unveiled at GTC to battle again
Huang addressed this instantly. Nvidia unveiled Dynamo, an clever inference engine that dynamically routes workloads throughout GPUs, ASICs, and CPUs relying on which is most effective. The corporate additionally confirmed that Vera Rubin is in full manufacturing and on monitor for a second-half 2026 ramp, per Nvidia.
Extra Nvidia:
- Nvidia inventory will get main actuality examine on ‘$100B’ quantity
- Nvidia CEO delivers blunt 7-word rebuttal on software program shares
- Financial institution of America resets Nvidia value goal after earnings
The Rubin platform delivers a 10x discount in inference token prices in comparison with Blackwell, which might make the economics of AI deployment considerably cheaper for each firm operating fashions at scale.
Wolfe Analysis analyst Chris Caso referred to as the Rubin ramp the important thing catalyst. He famous that Rubin delivers a 5x inference enchancment over Blackwell and that Nvidia’s up to date outlook implies at the very least $40 billion in upside to present consensus income estimates for calendar yr 2026, per Investing.com.
“Blackwell is now ramping fully, and Rubin is on time for 2H26 ramp,” Caso wrote, including that resumed H200 shipments to China might add additional upside past that determine.
The place the main analysts stand
Financial institution of America’s Vivek Arya reaffirmed his Purchase ranking and $300 value goal after the keynote. BofA sees greater than $1 trillion in information heart gross sales visibility for 2025 by 2027, up from a previous estimate of $500 billion by 2026.
Goldman Sachs, additionally sustaining a Purchase with a $250 goal, referred to as the $1 trillion determine one thing that “can help resolve investor concerns around peak CapEx in 2026,” per Investing.com. Citi and JPMorgan each reaffirmed $300 targets.
Wolfe Analysis reiterated its Outperform ranking with a $275 value goal post-GTC, noting the income disclosure suggests significant upside to 2027 estimates. Wells Fargo, at $265, mentioned Nvidia’s up to date order visibility merely “beats the bogey.”
The place analysts stand post-GTC
- Financial institution of America: Purchase, $300 goal, $1 trillion-plus information heart visibility confirmed
- Goldman Sachs: Purchase, $250 goal, referred to as $1 trillion outlook “above the Street”
- Wolfe Analysis: Outperform, $275 goal, $40 billion upside to CY26 consensus
- Citi: Purchase, $300 goal, tens of billions in extra upside not but in estimates
- JPMorgan: Purchase, $300 goal, $50-70 billion upside to 2026-2027 information heart consensus
- Wells Fargo: Obese, $265 goal, order visibility beats expectations
Why the inventory barely moved
Regardless of the bullish analyst refrain, NVDA has traded in a slim $180-$190 vary for months. The muted response to GTC continues a sample. After Nvidia’s October 2025 GTC occasion, the inventory fell 2% the next day.
Nagle/Bloomberg by way of Getty Pictures
The problem is timing. A lot of the upside analysts are projecting is backloaded into the second half of 2026, with Rubin not meaningfully contributing to income till late within the yr.
There’s additionally the China wildcard. Nvidia’s Q1 steering of $78 billion explicitly excludes China information heart income, following the export restrictions that pressured a $4.5 billion H20 cost earlier in 2025. Analysts estimate resumed China gross sales might add $10 billion to $15 billion to calendar yr 2026 income, however that relies upon totally on coverage selections exterior Nvidia’s management.
The execution danger is actual
Nvidia is buying and selling at roughly 38 occasions trailing earnings, a big premium to the broader semiconductor sector. At that valuation, execution danger leaves nearly no room for error
Any stumble within the Rubin ramp, additional China restrictions, or significant inference share loss might compress the inventory sharply. Amazon, Google, and Microsoft are all deploying extra customized silicon for inference workloads, which might restrict Nvidia’s addressable market whilst the general inference alternative grows.
Nvidia’s counter is Dynamo, designed to behave as a software program routing layer on prime of heterogeneous infrastructure, protecting Nvidia related even when a competitor’s {hardware} is within the rack. Whether or not that technique holds as customized silicon matures is the query traders are ready to see answered.
The consensus on Wall Road is evident. The roadmap is credible. The $1 trillion goal is actual. The inventory will observe when the income does.
Associated: Nvidia bull drops stunning tackle upside
