The story keeps changing, but one thing is for sure. There will be one more AI vendor who will need tens of thousands of new GPUs.
As the Loony Tunes saga continues, thanks to the Open.ai Board of Directors (”Th.. Th.. Th.. That’s all, Folks!”), a ton of meetings in Silicon Valley and Seattle worked through the weekend and late into the night trying to figure out a) “how can we capitalize on this self-inflicted near-fatal wound” and b) “what new threats might we face as a result?”
I won’t rehash the premise here. You already know the drill unless you have been on a beach without internet connectivity.
Here are some short and sweet possible impacts so you can get back to X.
Sam wins.
He will finally be able to get some equity!
Microsoft wins.
Microsoft will win by a lot unless Open.ai does a swift (in the next 2 hours?) about-face to try to keep their employees on board. First, they get the God-Pod of Open.ai leadership. Second, hundreds of employees will follow ASAP while they are still mad as hell. Third, Microsoft and Azure will now have carte-blanche to take all the weights of GPT-4 (and -5?) and whatever other IP they can claim and tell Sam, “Go for it.” Fourth, Azure Cloud is about to place a massive order for more accelerators, increasing market share against Amazon AWS and Google Cloud Compute. Finally, some of those orders will be fore Microsoft’s new Maia accelerator!
NVIDIA could win if...
NVIDIA has already increased the outlook for GPUs astronomically based on demand for Hopper and GraceHopper chips with associated CoSWoS capacity needed for HBM. NVIDIA wins if they can redirect a few tens of thousands to the new Microsoft AI Subsidiary (in Azure) without losing customers to AMD.
AMD could win with more CPUs, but GPUs?
AMD has a good share of CPUs attached to the NVIDIA GPUs mentioned above and could get a boost in orders to provision them with their leading Epyc CPUs. As for the upcoming Instinct MI300, we have already heard that Open.ai is excited about the prospect, and that excitement probably carries over with the team heading to Microsoft. AMD could be the biggest winner if they can get allocations from TSMC.
Intel will probably win with Xeon, but Gaudi?
In the same vein, Intel will sell a lot more Xeons. However, Gaudi is facing a challenging transition to Gaudi3, so Microsoft may not jump at the chance of buying tens of thousands of Gauidi2. While Gaudi3 looks very promising, I suspect it is too late.
Open.ai loses.
Duh.
Google loses.
Any gain by Microsoft in AI talent is Google’s loss unless they can outhire Microsoft.
AI Safety loses unless…
Microsoft does a better job than Open.ai in balancing speed and safety in AI. Keeping my fingers crossed on this one!
Disclosures: This article expresses the author’s opinions and
should not be taken as advice to purchase from or invest in the companies mentioned. Cambrian AI Research is fortunate to have many, if not most, semiconductor firms as our clients, including Blaize, BrainChip, CadenceDesign, Cerebras, D-Matrix, Eliyan, Esperanto, FuriosaAI, Graphcore, GML, IBM, Intel, Mythic, NVIDIA, Qualcomm Technologies, Si-Five, SiMa.ai, Synopsys, Ventana Microsystems and Tenstorrent. We have no investment positions in any of the companies mentioned in this article and do not plan to initiate any in the near future. For more information, please visit our website at https://cambrian-AI.com.
Read the full article here
