Forging The Future with Chris Howard

The episode contains a handful of substantive ideas - token cost subsidization, compute-near-memory biomimicry, planning for inference architecture, and Microchip's multicast offloading feature - but these are interspersed with extended filler: humanoid robot philosophy, horse-and-carriage metaphors, Matrix/Borg riffs, and the host's extended solo musings. Useful insight per minute is moderate at best.

“focusing on standardization of the training tasks which are the most expensive and planning for inference. Inference will be considerably cheaper than training”

“we've invented this multicast feature where the GPU if it needs to send data packets to 10 different storage locations today, uh, or traditionally a GPU would have to send 10 different instructions to 10 different destinations. We've created a technology that allows the GPU to delegate that to the switch”

The AI-eras taxonomy (machine learning → training → inference → AGI) and the biomimicry compute-near-memory framing are the freshest angles, but neither is genuinely contrarian or first-principles. The Uber/taxi token-economics analogy is mildly clever but generic, and most other frameworks (physical AI, agentic AI, edge inference) recycle Jensen Huang talking points rather than adding new perspective.

“if you look at every living organism on planet Earth, from a jellyfish to ah, a cow to a human and everything in between, they all share a compute near memory or compute in memory architecture”

“So we are going through a very similar replacement type business model right now as what the world saw with taxi cabs”

McCarson is a genuine practitioner - almost 25 years at Intel, now leading a data center business turnaround at Microchip - and speaks with the credibility of someone who has shipped silicon and managed P&Ls. He is not a career podcast guest or pure thought-leader, but he is also not a CEO of a marquee company, and the episode partly functions as Microchip brand promotion rather than fully arms-length expertise.

“I joined Microchip at the beginning of 2025 and uh, that was after a long stint, um, mostly over my career. I've worked at a few companies but spent almost a quarter of a century at Intel Corporation”

“In my career, turning around businesses is something that I've done serially uh, over many, uh, different endeavors”

There are genuine specifics - TSMC 3nm, 160 lanes at 64 billion transfers/second, 10 teratransfers per second at 80 watts, token costs allegedly 90%+ subsidized, narrow AGI four years out - but most quantitative claims are hedged ('I've seen several articles suggesting,' 'far north of 95%') or delivered without sourcing. Named examples like Amazon Alexa are illustrative but well-worn, and many strategic claims remain at the level of analogy rather than data.

“this switch device here uh, you know, in particular has um, uh, it's a switch that has 160 different lanes, um, that can each handle that parallel traffic. But each of those lanes can do ah, 64 billion transfers per second”

“I've seen several articles suggesting we're getting a 90 plus percent discount right now on the cost”

The host asks a few genuinely useful operational questions (token budget management, how to architect for inference, energy constraints) but frequently derails into personal riffs on religion, the Borg, and Doordash robots without redirecting to actionable follow-up. Bold claims - AGI in four years, 90% token subsidies, space data centers - go unchallenged, and there is a mid-episode subscription pitch that breaks flow entirely.

“The fourth era is that Robopocalypse?”

“And to be clear, those PhDs were your AI agents, right?”