By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Sourcery with Molly O'SheaBlackRock's Tony Kim on AI's Next Winners? Chips, Memory, Robotics & Quantum
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
AI flips value to compute: data centers, memory, robotics, 2030
- Tony Kim argues AI has flipped tech from software-centric value capture to compute- and hardware-centric economics, forcing investors to “rethink everything.”
- He says AI’s compute demands require a complete rebuild of data centers, where shrinking transmission distances and rising density make bandwidth, heat, and power the governing constraints.
- He highlights a shift from copper interconnects to optics and from compute-only focus to rising memory importance, calling out a looming “RAMpocalypse” driven by memory intensity and slow capacity build times.
- He outlines an investment approach that concentrates on near-term AI supply-chain winners while also placing smaller bets on frontier technologies that he believes converge around 2030.
- He points to robotics as the next major embodiment of AI—split between “brain” (models) and “body” (manufacturing)—and expects major activity from China while emphasizing companionship/social robots as an underappreciated market.
IDEAS WORTH REMEMBERING
5 ideasAI has moved tech’s center of gravity from software margins to compute-intensive infrastructure.
He described a shift from a “software-centric world to a compute-centric world,” where AI has made the base layer of compute dramatically more valuable and central to company economics.
AI is forcing a ground-up redesign of the data center.
He argued that older cloud-era data centers were built for CPUs and hard drives and are now being replaced by “alien” AI-era facilities driven by power density, heat, bandwidth, and new physical constraints.
The laws of physics are reshaping connectivity: copper gives way to optics as distance collapses.
He said the industry is moving “from a regime of copper to a regime of light” as distances shrink from kilometers to meters to centimeters to millimeters, pushing optical interconnects and new architectures.
Memory is becoming a primary bottleneck, and supply can’t adjust quickly enough.
He tied “RAMpocalypse” to models becoming more memory-intensive (more like a brain) and to the supply-duration mismatch where new memory capacity takes “three or four years” to build while shortages are happening now.
Model–silicon co-design is becoming a core competitive strategy in AI.
He said leading players are increasingly pursuing tight “co-design,” where silicon is optimized for model requirements and model specs inform compute design, citing this as a key direction for the stack.
WORDS WORTH SAVING
5 quotesSo the base layer of compute went up 10,000X. $10,000 server is a million-dollar server.
— Tony Kim
The trillion dollars of CapEx this year and the $10 trillion over the next five years that are coming is to move data centimeters and millimeters. That's AI.
— Tony Kim
I think the primacy of memory will become even more important.
— Tony Kim
People always said chips are a commodity, but yet they have the highest profitability of any company in the world, of the sector, is chip companies.
— Tony Kim
Will you still be cool in five years?
— Tony Kim
SPEAKERS
Molly O'Shea
hostHost of Sourcery with Molly O'Shea, interviewing tech and AI leaders and reading show sponsor segments.
Tony Kim
guestBlackRock global technology investment leader discussing AI infrastructure themes such as chips, memory, robotics, and quantum.
Molly O'Shea
hostHost of Sourcery with Molly O'Shea, handling closing questions, sponsor reads, and episode outros.
High quality AI-generated summary created from speaker-labeled transcript.
