For much of the AI boom, there have been whispers about Nvidia’s frenzied dealmaking. The chipmaker bolstered the market by pumping money into dozens of AI startups, many of which rely on Nvidia’s graphics processing units to develop and run their models. OpenAI, to a lesser degree, also invested in startups, some of which built services on top of its AI models. But as tech firms have entered a more costly phase of AI development, the scale of the deals involving these two companies has grown substantially, making it harder to ignore.

Now that all the pieces are in place, here is the economic nexus of semi/genAI that particularly interests me:

If model providers make inference much more efficient, then they will not use enough computing power to consume all that is brought to market by the semiconductor industry. If this happens, it will trigger a downward cycle in this industry, significantly slowing down the production of new hardware and possibly having significant global economic and financial repercussions.
If model providers do not make their inference processes more efficient, they will not be able to structurally reduce their marginal costs and, failing to achieve the desired profitability, will resort to the usual means (advertising, tiered subscriptions), which will slow down adoption.
If adoption slows down, model providers will struggle to achieve profitability (with the exception of those with captive markets), their demand for computing power will weaken, and the semiconductor industry will produce excess capacity and enter a downward cycle, taking part of the AI industry with it.

So, the central issue linking today’s semiconductor industry and genAI model providers is how to define how much efficiency gains are enough. Jokingly, we could call this ‘inference inefficiency optimum’.

on pc gaming, violence, hardware

stats about the machines used by Steam subscribers

Monthy 3Digest is the project of iXBT.com and Digit-Life.com dedicated to informing you about the performance of a large number of graphics cards under Windows XP.