#Gate广场中秋团圆局 #AMD预计2030年AI市场达2万亿美元 At the Citi Global TMT Conference on September 8, AMD doubled its 2030 total addressable market (TAM) for AI from approximately $1 trillion to $2 trillion (the CFO verbally mentioned as much as $3 trillion), including $1.4 trillion for AI accelerators ($200 billion in 2025, with a 45% CAGR) and $220 billion for server CPUs. It also announced that its Helios rack systems had received excess orders from customers including Meta, while OpenAI is expected to deploy them at scale. AMD rose 7% intraday and closed up approximately 6%, while CLSA raised its price target to $710.

How should this be interpreted: take it with a grain of salt
The TAM increase is a typical “sell-side narrative,” so its reference value should be discounted by half.
The shovel seller is telling you that “more and more people will mine gold”—of course, this is being said in line with its own interests. The direct effect of AMD raising its TAM is to find an anchor for its own valuation. $1 trillion → $2 trillion → $3 trillion—a three-step jump within a few months (it only said $1.4 trillion for accelerators in July). The faster the definition jumps, the more this shows “expectations management” rather than “scientific estimation.”
But the excess orders for Helios are real, and this is the substantive part of this rally. Meta, plus two unnamed AI labs, have demand exceeding their initial purchase agreements; OpenAI is deploying at scale; and the number of inference tokens per dollar is 30% higher than competitors’—these are hard evidence from the demand side, mutually corroborated by NVDA’s better-than-expected Q3 guidance and Micron’s HBM sold out through the end of 2026. The AI capex cycle is still expanding; this cannot be explained by a bubble narrative alone.

Will the AI market continue to rise? The answer has two layers
At the industry level: most likely, yes. The basis for this judgment is not AMD’s presentation, but three independent chains of evidence:
①NVDA’s FY28 growth guidance of 70%+ and quarterly data-center revenue of $89 billion;
②Micron’s HBM sold out, with MU’s earnings expected revenue at $50.4 billion;
③Microsoft/Meta/OpenAI are still increasing their capital expenditures. Supply-side constraints (HBM and advanced packaging) remain bottlenecks, indicating that demand has not been fully satisfied.
At the stock-price level: most of the gains have already been priced in over the short term, and the stock is extremely sensitive to interest rates.
Pay attention to two signals: first, after AMD rose 7% in a single day, it fell 2.5% the next day to $508, indicating that the market’s reaction to the TAM increase was one-off and did not generate sustained buying;
Second, Goldman Sachs expects the September FOMC to raise rates by 25 basis points, and interest rates are the biggest enemy of AI stocks with high valuations. Morgan Stanley also warned that the growth rate of AI infrastructure investment will slow in 2027–28, with application spending taking over—meaning the sweet spot for “infrastructure stocks” is limited.

Conclusion

The AI industry’s upward cycle is not over, and the “peak” theory has been disproven once again; but “the AI market will continue to rise” does not mean “chasing AI stocks at current levels will make money.”
The short-term rhythm is this: the September FOMC is the watershed—if the rate hike is implemented without signaling a series of further hikes, as Goldman Sachs expects, AI stocks will most likely rebound after the negative news is fully priced in; if a hawkish signal is delivered, high-valuation chip stocks will be hit first.
Trading reference: do not chase AMD around $508–510 (it has just been boosted by the TAM news); wait for the FOMC outcome and then assess the direction; $470–480 (the previous consolidation range) is a better entry zone. The real validation will come during the earnings season from late October to November—NVDA, AMD, and MU’s three earnings reports will tell you whether AI capex is “just beginning” or “overestimated.”

When TAM jumps from $1 trillion to $2 trillion and then to $3 trillion, and the people shouting about it become increasingly excited, that is precisely when you should check your position discipline—the AI trade is the most certain theme of this cycle, and also the theme most likely to make people forget to use stop-losses. $AMD