korrents

Doug O'Laughlin

@doug-olaughlin · 12 positions · 1 change of mind

Semiconductor analyst, author of the Fabricated Knowledge newsletter and later a partner at SemiAnalysis, writing on the capital cycle in chips: capex, memory supply, packaging, and where the gluts come from.

Doug O'Laughlin did not write this page.

We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.

  1. Our reading

    Cheaper AI will be used more in the long run and still strand the hardware built for it in the short run, exactly as unlit fibre did after 2001.

    Their wordsAnd so, while the Jevons paradox case was 100% correct in the long run, 97% of the fiber laid in 2001 was unlit.

    ↗DeepSeek: Is this Jevon's Cope?fabricatedknowledge.com

    Interesting
  2. 2 weeks earlier

  3. Our reading

    Every semiconductor shortage is followed by a glut, because capacity arrives in lumps long after the price signal that ordered it.

    Their wordsA glut follows every semiconductor shortage.

    ↗Capital Cycles and AIfabricatedknowledge.com 1st of 3 in this piece

    semiconductors

    Interesting

  5. Our reading

    The AI build-out is not yet a bubble, because it lacks the self-reinforcing feedback loop that drove the railway and telecom manias.

    Their wordsWhat prevents this from being a broad-based bubble is that we are missing a critical feedback loop from the announcement's impact, like network growth or land value.

    ↗Capital Cycles and AIfabricatedknowledge.com 2nd of 3 in this piece

    Interesting

  7. Our reading

    If the AI build-out does become a bubble, power and data centres are the assets that will carry it.

    Their wordsIn today’s world, I think the increase in self-fulfilling asset prices will be due to power and data centers.

    ↗Capital Cycles and AIfabricatedknowledge.com 3rd of 3 in this piece

    data centers

    Interesting
  8. 9 months earlier

  9. Our reading

    Moore's Law is a fractal: the same squeeze that shrank the transistor is now being applied to the rack.

    Their wordsMoore’s Law is a fractal, and the principles that apply to nanometers apply to racks.

    ↗The Data Center is the New Compute Unit: Nvidia's Vision for System-Level Scalingfabricatedknowledge.com 1st of 3 in this piece

    Interesting

  11. Our reading

    The copper backplane inside a rack is the new advanced packaging, and scaling it is where the next generation of performance comes from.

    Their wordsThe copper backplane in the data center rack is effectively the new advanced packaging in the system-level Moore’s Law race.

    ↗The Data Center is the New Compute Unit: Nvidia's Vision for System-Level Scalingfabricatedknowledge.com 2nd of 3 in this piece

    data centers

    Interesting

  13. Our reading

    Air cooling has run out of headroom for AI racks, so liquid cooling is now a requirement rather than an upgrade.

    Their wordsThe critical change that makes this all possible is liquid cooling.

    ↗The Data Center is the New Compute Unit: Nvidia's Vision for System-Level Scalingfabricatedknowledge.com 3rd of 3 in this piece

    Interesting
  14. 5 weeks earlier

  15. Our reading

    High-bandwidth memory costs roughly twice the capital per bit that ordinary DRAM does, because its dies are twice the size.

    Their wordsHBM has the unique dynamic of costing a lot more than traditional DRAM, and the dynamic 2x higher capex is driven by the twice as large die size for HBM.

    ↗Why HBM is the Hottest Thing in Memoryfabricatedknowledge.com 1st of 2 in this piece

    Interesting

  17. Our reading

    The rush to build high-bandwidth memory starves ordinary DRAM and NAND of capacity, so AI demand tightens the memory markets it never touches.

    Their wordsThey will shift what cash they generate to HBM capacity, which will keep supply growth in DRAM and NAND low.

    ↗Why HBM is the Hottest Thing in Memoryfabricatedknowledge.com 2nd of 2 in this piece

    Interesting
  18. 2 years earlier

  19. Our reading

    Chipmakers are consolidating into a few strategically important firms, so a fab is now a geopolitical asset rather than an industrial one.

    Their wordsFinally, not just old chips and new chips, but the companies that make the chips (semiconductor fabs) are becoming more consolidated and more strategic.

    ↗The Rising Tide of Semiconductor Costfabricatedknowledge.com 1st of 3 in this piece

    semiconductors

    Interesting

  21. Our reading

    Chips are getting more expensive mainly because they take more process steps, not because the tools got dearer.

    Their wordsThe primary driver is not only the rising costs of tools such as EUV, but the rising number of steps to make a chip.

    ↗The Rising Tide of Semiconductor Costfabricatedknowledge.com 2nd of 3 in this piece

    Interesting

  23. Our reading

    TSMC can pass its rising costs on because at the leading edge its customers have nowhere else to go.

    Their wordsGiven the concentration of fabrication power at TSMC, customers really don’t have another option.

    ↗The Rising Tide of Semiconductor Costfabricatedknowledge.com 4th of 3 in this piece

    semiconductors

    Interesting