#NvidiaMarketCapBackAbove5.4T

Yes — the AI-compute thesis is still very much intact, but from here the question changes from “Is AI demand real?” to “How long can Nvidia sustain extraordinary growth while already being worth ~$5.45T?”

Nvidia’s FY27 Q2 was extraordinary: revenue hit $96.2B, +106% ​#NvidiaMarketCapBackAbove5.4T
Yes — the AI-compute thesis is still very much intact, but from here the question changes from “Is AI demand real?” to “How long can Nvidia sustain extraordinary growth while already being worth ~$5.45T?”

$NVDA Nvidia’s FY27 Q2 was extraordinary: revenue hit $96.2B, +106% YoY, data-center revenue reached roughly $89B, +117%, and management guided Q3 revenue to $108B ±2%. Gross margin was 75%.

How far can the rally go?

I’d frame it in scenarios rather than one target:

Base bullish $6–7T Rubin ramps cleanly + AI capex remains elevated

Strong bull $8–10T Inference becomes a massive recurring compute market

Super bull $12T+ 60–70% growth persists well beyond FY28 + margins stay exceptional

Bear case $3.5–4.5T Hyperscaler capex slows, custom ASICs gain share, or AI monetization disappoints

The really important development is Vera Rubin. This isn’t merely another faster GPU generation. Nvidia is positioning Rubin around inference economics and agentic AI, where the amount of compute consumed can grow dramatically as AI applications become more autonomous. Rubin is already in full production.

And that’s where Huang’s “compute is revenue” argument becomes powerful.

If AI inference generates economically valuable tokens, then every additional query/agent/task creates demand for compute. That means Nvidia isn’t necessarily approaching a traditional “companies buy enough GPUs and then stop” cycle. The potential demand curve becomes tied to AI usage, not merely data-center construction.

But there’s a valuation trap here

At $5.4T, Nvidia doesn’t need to merely remain dominant. It has to keep surprising upward.

The market is already discounting extraordinary growth. One analysis estimates FY28 revenue could approach ~$450B if the current growth trajectory persists.

So I’d watch four things above everything else:

1. Hyperscaler capex → still accelerating?
Microsoft, Amazon, Google, Meta and the AI-cloud providers are effectively the fuel tank.

2. Inference → becoming bigger than training?
This is probably the biggest long-term catalyst. If agents cause inference workloads to explode, the compute market can expand even while training growth eventually normalizes.

3. Rubin → does it increase Nvidia’s customers’ ROI?
If Rubin substantially lowers cost/token, it could paradoxically increase demand: cheaper inference makes more AI applications economically viable.

4. Nvidia’s growth rate → 70% vs. 40–50%.
At $5.4T, that difference becomes enormous. A deceleration to 40–45% could produce a major multiple compression even if Nvidia continues growing spectacularly.

My take

I wouldn’t call $5.4T the end of the AI rally.

I’d call it the point where the upside becomes increasingly dependent on the second phase of AI.

The first phase was:

Build the infrastructure.

The next phase is:

Use the infrastructure continuously to generate revenue.

If that transition succeeds, $8–10T Nvidia is not an absurd scenario over the next several years. But getting from $5.4T to $10T requires roughly another 85% increase in equity value, so earnings and cash flows have to grow enormously alongside it.

The biggest risk isn’t AMD suddenly “beating Nvidia.” It’s AI compute demand growing more slowly than the valuation assumes — particularly if hyperscalers discover that their AI investments aren’t generating sufficient returns, or if custom silicon takes a much larger share.

So my bull/bear dividing line is simple:

AI usage ↑ → inference ↑ → compute demand ↑ → Nvidia earnings ↑ → rally can continue.

AI capex ↑ but AI revenue disappoints → ROI falls → capex eventually slows → Nvidia multiple contracts.

Right now, the evidence from the latest quarter is overwhelmingly on the first side.

My rough probability-weighted view: the next major leg could plausibly take Nvidia toward $6–7T first, with $8–10T becoming realistic if Rubin + inference produce another 2–3 years of exceptional growth. $12T+ requires something closer to an AI-compute supercycle than simply “another great Nvidia cycle.”

And that distinction is crucial.YoY, data-center revenue reached roughly $89B, +117%, and management guided Q3 revenue to $108B ±2%. Gross margin was 75%.

How far can the rally go?

I’d frame it in scenarios rather than one target:

Base bullish $6–7T Rubin ramps cleanly + AI capex remains elevated

Strong bull $8–10T Inference becomes a massive recurring compute market

Super bull $12T+ 60–70% growth persists well beyond FY28 + margins stay exceptional

Bear case $3.5–4.5T Hyperscaler capex slows, custom ASICs gain share, or AI monetization disappoints

The really important development is Vera Rubin. This isn’t merely another faster GPU generation. Nvidia is positioning Rubin around inference economics and agentic AI, where the amount of compute consumed can grow dramatically as AI applications become more autonomous. Rubin is already in full production.

And that’s where Huang’s “compute is revenue” argument becomes powerful.

If AI inference generates economically valuable tokens, then every additional query/agent/task creates demand for compute. That means Nvidia isn’t necessarily approaching a traditional “companies buy enough GPUs and then stop” cycle. The potential demand curve becomes tied to AI usage, not merely data-center construction.

But there’s a valuation trap here

At $5.4T, Nvidia doesn’t need to merely remain dominant. It has to keep surprising upward.

The market is already discounting extraordinary growth. One analysis estimates FY28 revenue could approach ~$450B if the current growth trajectory persists.

So I’d watch four things above everything else:

1. Hyperscaler capex → still accelerating?
Microsoft, Amazon, Google, Meta and the AI-cloud providers are effectively the fuel tank.

2. Inference → becoming bigger than training?
This is probably the biggest long-term catalyst. If agents cause inference workloads to explode, the compute market can expand even while training growth eventually normalizes.

3. Rubin → does it increase Nvidia’s customers’ ROI?
If Rubin substantially lowers cost/token, it could paradoxically increase demand: cheaper inference makes more AI applications economically viable.

4. Nvidia’s growth rate → 70% vs. 40–50%.
At $5.4T, that difference becomes enormous. A deceleration to 40–45% could produce a major multiple compression even if Nvidia continues growing spectacularly.

My take

I wouldn’t call $5.4T the end of the AI rally.

I’d call it the point where the upside becomes increasingly dependent on the second phase of AI.

The first phase was:

Build the infrastructure.

The next phase is:

Use the infrastructure continuously to generate revenue.

If that transition succeeds, $8–10T Nvidia is not an absurd scenario over the next several years. But getting from $5.4T to $10T requires roughly another 85% increase in equity value, so earnings and cash flows have to grow enormously alongside it.

The biggest risk isn’t AMD suddenly “beating Nvidia.” It’s AI compute demand growing more slowly than the valuation assumes — particularly if hyperscalers discover that their AI investments aren’t generating sufficient returns, or if custom silicon takes a much larger share.

So my bull/bear dividing line is simple:

AI usage ↑ → inference ↑ → compute demand ↑ → Nvidia earnings ↑ → rally can continue.

AI capex ↑ but AI revenue disappoints → ROI falls → capex eventually slows → Nvidia multiple contracts.

Right now, the evidence from the latest quarter is overwhelmingly on the first side.

My rough probability-weighted view: the next major leg could plausibly take Nvidia toward $6–7T first, with $8–10T becoming realistic if Rubin + inference produce another 2–3 years of exceptional growth. $12T+ requires something closer to an AI-compute supercycle than simply “another great Nvidia cycle.”

And that distinction is crucial.

$NVDA ‌