Move Over, Nvidia: Why AMD’s Latest Acquisition Could Redefine Direct-Silicon AI Inference

Just as investors were busy digesting that massive $500 billion Wall Street deal and the risks they entail if there is an AI bubble, Nvidia (NASDAQ:NVDA | NVDA Price Prediction) went on to back an OpenAI data center project with a whopping $105 billion.

The circular deals and financing moves are coming in fast, with this latest move tying right back to Sam Altman’s OpenAI, which has been a source of much skepticism and even a bit of hesitancy of late following a key person’s departure, chief revenue officer Denise Dresser, as well as growing doubts about the company’s financials as a market-moving IPO looms.

Indeed, Nvidia is making massive deals with breakneck speed. And while each new move, especially those focused on frontier labs or data centers, will just make the bears more bearish, the bulls also stand to get more bullish, as the stakes take things up several more notches. Not to doubt Jensen Huang’s ability to spot alpha across the AI waters, but I do think that as the deals stack up, some of the bulls might grow content with taking a few chips off the table.

Advanced Micro Devices just scooped up a rising tech star

Either way, Advanced Micro Devices (NASDAQ:AMD) and Lisa Su are racing to gain any edge they can get as the silicon wars heat up and AI agents look to take off. It’s easy to overlook or even outright dismiss Advanced Micro Devices’ latest acquisition of a little-known Canadian startup named Taalas.

Indeed, it was a rather small scoop-up, but I do think it shines a bright light on model-specific silicon, which may very well represent one of the next big steps for AI chips as inference takes away a bit of the luster from general-purpose GPUs.

Undoubtedly, Taalas’ impressive tech, which strives to program AI models right onto the silicon itself, could help Advanced Micro Devices power through the memory and power walls. Indeed, there’s a lot more to gain by bringing the Taalas team aboard Lisa Su’s empire, as efficiency gains and, perhaps more importantly, breaking through AI bottlenecks look to be the name of the game.

Time will tell how much a Toronto-based startup moves the needle at a $826 billion behemoth. But, in my view, Advanced Micro Devices is getting bang for its buck as it onboards new talent who are focused on innovating their way past constraints that seem to be holding most other firms back.

While it’s too early to tell how the “phenomenal team working at the bleeding edge of AI inference,” as Lisa Su put it, will help change Advanced Micro Devices’ trajectory from here, I do think the firm’s unique, differentiated approach of “building the hardware around the model” could be a source of breakthrough surprises.

Any way you look at it, it will be interesting to see how that strategy scales as Advanced Micro Devices looks to put its money to work on acquisitions in a way that could change the game rather than just add to circularity with considerable sums. I’m a huge fan of Advanced Micro Devices’ latest deal and must say that Taalas is an acquisition that genuinely makes me excited for what comes next. I can’t say the same about Nvidia’s latest financing deals.

The bottom line

For now, it’s hard to say where direct-silicon AI inference goes from here. It adds a whole new layer of efficiencies that might just be key to running next-generation agentic AI in an era where scaling up just won’t do it with all these memory and energy hurdles ahead.

For Advanced Micro Devices, though, the Taalas deal bolsters the portfolio and might allow a breakthrough that grants Lisa Su’s firm an edge that helps it close a bit of the gap with Nvidia by using a different kind of playbook.

Contact [email protected] for any questions or corrections.