AI: Weekly Summary. RTZ #772

Michael Parekh AI: Reset to Zero

3 weeks ago

5 MIN READ

Meta AI Talent Gold Rush: Meta continued to accelerate its moves to revamp its AI team and strategy with a range of hires across the industry, with meaningful impact on AI Talent markets. In addition to its ‘acqui-hire’ of Scale AI for $14 billion, Meta also is is bringing on VC and AI founders Nat Friedman and Daniel Gross for large sums. In addition, Meta founder/CEO Mark Zuckerberg is now on an accelerated mission to AI Superintelligence, with Alexandr Wang now leading the AI efforts with the help of Nat and Daniel in the top management ranks. Despite OpenAI’s claims and efforts, Meta has seen some success luring top AI research talent from OpenAI and others. These moves are likely more par for the AI course for this year and beyond. More here.

China’s role in Global AI Talent Market: Meta’s AI Talent moves also have wider implications globally, given that half of the global AI Research talent is in and from China. It’s a point well made by Nvidia founder/CEO Jensen Huang, who is also laser focused on that opportunity and reality. It’s notable that even the most recent Meta hires in the US of AI research talent from OpenAI and others are US trained Chinese candidates, many who’ve gone on to successful AI research careers in the US. Both in academia and business. Additionally, China remains a key global leader in open source LLM AIs, and Chinese startups there continue to innovate and iterate despite US tech trade curbs and tariffs. So it’s important to frame the AI Talent wars as a global one, with China very much in the picture. More here.

AI Industry Impact of Meta AI Talent Moves: At this mid-point in 2025, the AI Table Stakes have risen across the board beyond AI chips, data centers and other infrastructure, to now include AI research and other talent. Meta’s moves simply accelerate a trend already in the making. We have seen this in other tech cycles, most recently in the mobile wave. The implications here go beyond the tech and AI industries, and include companies across most industries also ramping up AI capable talent in their ranks. It’s not the $100 million AI research hires from OpenAI, but AI talent up and down the AI tech stack, at all levels. It’s a reality that will drive the industry for the rest of this year and beyond. It has notable implications for everyone of the ‘Mag 7’ focused on AI with billions to invest. It’s a game of AI Models and Talent musical chairs. More here.

Rising Challenge of AI Variable Cost driven Pricing: The AI/Tech industry continues to see the high class problem of managing its variable costs as customers embrace a wide range of LLM AI models for an accelerating array of applications and services. Even though the models continue to be more cost efficient to run from a software and hardware perspective, the higher inference use in particular means higher variable costs for customers. Some of this is being addressed by LLM AI companies charging higher prices both in terms of APIs and subscription tiers, with OpenAI, Perplexity and others recently going to $200/month tiers and possibly higher. This reality will likely be the norm as AI Reasoning and Ai Agents become more useful for customers to deploy. More here.

Microsoft’s AI Chip Challenges: Microsoft is running into challenges with its own AI chips as a backup to Nvidia for its core AI needs at Azure and elsewhere. The delays range from six months to longer across a range of chips planned by the company. Microsoft and other major Nvidia cloud customers like Amazon, Google, Meta, OpenAI et al are all engaged in similar ‘Frenemies’ efforts to build their own AI chips for both training and inference needs. Microsoft’s issues are an example of some of the technical and other challenges in the multi-year path to these alternatives, Microsoft remains one of Nvidia’s top customers, along with both companies being close to $4 trillion market caps each. With Apple of course also being in those ranks, and with their own Ai chip efforts. More here.