What is NVIDIA falling for?

1. Supply Chain

First, here is a number just released today by Taiwan's Fubon Investment Consulting, the latest adjusted order volume for GB200+GB300, as well as the updated schedule.

The volume needs no explanation; everyone can understand it themselves. The schedule aligns with Hon Hai's latest statement: GB will enter MP large-scale production and shipment after the Lunar New Year in February. Four months later, GB300 will have a small batch, which we previously mentioned, giving GB200 a bit short of time...

Over the past month, we have repeatedly mentioned "a bunch of small issues," such as CoWoS-L, heat generation, copper cables, liquid leakage, etc. The overall yield improvement requires more time, and according to Jensen Huang's original words, "different customers and different configurations require time for system integration." Please note that this is the underlying reason, which is why there are so-called Microsoft's order transfers and NV's suggestion for customers to first buy B200 8-card HGX. At the same time, we have also uploaded the recent two small meetings of the CFO and IR, and the tone is relatively optimistic—"Blackwell is still likely to exceed Hopper in Q1," meaning that even if GB200 ramps up slowly, most of the pressure is actually on ODMs (just look at the stock prices of Hon Hai and Quanta...). Because under the Buy&sell model, NVIDIA confirms revenue when selling computing boards to Hon Hai, and the longer the cabinet integration takes, the longer Hon Hai bears the inventory cycle for NV. Therefore, NV will always find ways to ship chips to deliver qualified performance, but it is conceivable that the upside will not be too large, especially since there is also pressure on H...

2. Competing Narratives

The biggest change compared to the delay in August is: 1) the emergence of a competing narrative around ASICs; 2) changes in scaling law narratives and strengthening of reasoning narratives; 3) software and applications, BTC, Tesla/Waymo, all represent competing narratives from a funding perspective... especially for retail investors, there are more choices.

Most of this has been digested, except for the ASIC narrative and reasoning narrative, which have mutually strengthened over the past two weeks. Especially with Ilya's speech and Satya's model of "devaluation theory," which has shifted the narrative of model training towards reasoning. At this time, AVGO Broadcom's CEO came out to make bold claims, further igniting the sentiment that had already been heated by AWS Tr2 regarding ASICs. Subsequently, some began to question the TCO (total cost of ownership) of NVIDIA GPUs... For example, the Morgan Stanley ASIC in-depth report uploaded yesterday received the most feedback regarding that TCO comparison chart"Calculating to death" the GPU. Without elaborating on the details, the key point is that this represents a current market "view," which believes that ASICs indeed have potential.

Today, the industry leaders are intensely discussing the value of ASICs to CSPs, and whether if the future focus is on small model inference, GB200 still has such a significant advantage. These are very representative viewpoints recently.

3. Pure Garbage Time

As the year comes to an end, whether retail investors, ETFs, or institutions, everyone is balancing and adjusting their portfolios, looking at the recent Microsoft, Apple, especially Google. Taking profits by cutting high and switching to low, then going to celebrate Christmas... is a choice for many funds.

4. The Next Catalyst

Expectations for Jensen Huang's speech at CES in January should not be too high, as CES is essentially a consumer electronics show. There are also no expectations for performance at the end of February, as the Q4 buffer is not large. Then there’s the GTC in March, to see what Rubin can unfold. But personally, I feel that the marginal "surprise" will not be greater than the last GTC, where the focus was mainly on training and supply-side narratives, and the emergence of NVL72 was a groundbreaking moment. Now, simply relying on "a stronger supply side" might be a bit difficult. So what should NV be looking forward to?

The only thing that can move NV's 3 trillion is a new "computing narrative," a narrative that further opens up the imagination of computing, such as Agent or any super app. This will happen sooner or later; it's just a matter of time. o1 has set the stage, and OpenAI's 12 consecutive releases may open the second chapter, with other things possibly emerging from unknown places. Jensen Huang once said that when every entity in the world is running inference 24/7, AI will truly succeed. If this is the ultimate goal, then indeed, computing still has a long way to go, and so does NV. ASIC or GPU, software or hardware, are just small narratives under this larger narrative. AI cannot be stopped now, and humanity will also be inseparable from computing power in the future.

What should NV do? There is indeed short-term pressure; it’s a matter of what price to bottom out at, and you are welcome to join the discussion.

Source: Information Equality, Original Title: "What is NVIDIA Falling For"

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article align with their specific circumstances. Investing based on this is at your own risk