Teahose.
SIGN IN
NEW HERE — WHAT TEAHOSE DOES
We read the entire AI & tech firehose — so you don't have to.
PODPodcastsAll-In, No Priors, Acquired…
NEWNewslettersStratechery, Newcomer…
PAPPapersPhysical AI research
PHProduct Huntdaily launches
VCInvestor ScoutSequoia, a16z, Benchmark…
CLAUDE DISTILLS →
7 reads, 30 sec each — free, 6 AM ET.
+ a live graph of the companies, people & themes underneath.
HOME/COATUE/Chart of the Day
NEWS
// NEWSLETTER ISSUE
COATUE

Chart of the Day

DATE October 1, 2026SOURCE COATUEPARTICIPANTS COATUE MANAGEMENT
// SUMMARY

Coatue — Chart of the Day
Coatue — Chart of the Day

Coatue — Chart of the Day (2)
Coatue — Chart of the Day (2)

Coatue — Chart of the Day (3)
Coatue — Chart of the Day (3)

1. Key Themes

Theme: AI inference costs are collapsing at an unprecedented rate

AI cost deflation outpaces every prior transformative technology

The headline claim and chart place AI's price decline above electricity, lithium batteries, DNA sequencing, and compute, even though AI's decline has run for only a few years.

  • "The cost of AI has fallen faster than any other transformative technology in history!"
  • Chart title: "AI Inference Costs vs. Other Technologies Relative to Starting Price"

The pace is roughly 50% cheaper every quarter

The chart's callout quantifies the rate of decline since 2023, which compounds to a very large cumulative drop.

  • "AI costs have fallen ~50% per quarter since 2023!"

AI reached ~1 million x cheaper in about 4-5 years, versus decades for prior technologies

On the log-scale chart, AI's line (2021–Today) drops to roughly the "1 million x" level within about 5 years. DNA sequencing (2001–2022) reached a similar level only after about 20 years, and compute (1940–2001) took roughly 60 years to approach "1 trillion x." Lithium batteries (1991–2024) and electricity (1892–1973) fell far less over longer spans.

  • Chart labels: "AI 2021–Today," "DNA sequencing 2001–2022," "Compute 1940–2001," "Lithium batteries 1991–2024," "Electricity 1892–1973"
  • Axis: "Years since the start of price decline"

2. Contrarian Perspectives

AI cost deflation is steeper than the compute curve that powered the digital era

Many frame AI as an extension of Moore's Law-style compute declines. The chart suggests AI is compressing in years what compute took about six decades to achieve, which implies a faster adoption and disruption cycle than the internet or PC eras.

  • Evidence: Compute (1940–2001) ends near the "1 trillion x" line after ~60 years, while AI (2021–Today) is already near "1 million x" after roughly 5 years.
  • "The cost of AI has fallen faster than any other transformative technology in history!"
  • Caveat: the chart measures inference cost relative to starting price, so it does not capture capability-adjusted or total spend.

3. Companies Identified

Epoch AI

  • Description: Research organization tracking AI trends, including inference pricing.
  • Why mentioned: Cited as the data source for the chart.
  • Quote: "Source: Epoch AI. Note: All prices adjusted for inflation."

Coatue Management

  • Description: Investment firm and publisher of the "C:\Takes" newsletter.
  • Why mentioned: Author of the chart and the thesis on AI cost deflation.
  • Quote: "See more C:\Takes"

4. People Identified

None named in the article.

5. Operating Insights

Plan product economics around rapidly falling inference costs

With costs halving every quarter, features that are uneconomical today may be viable within a few quarters. Operators should design roadmaps and pricing with the expectation that unit costs will drop sharply.

  • "AI costs have fallen ~50% per quarter since 2023!"

Don't anchor to historical technology-adoption timelines

Prior cost curves (electricity, batteries, DNA sequencing) played out over decades, but AI's has played out over years. Competitive windows, and the time to build moats before costs commoditize, are likely much shorter.

  • "The cost of AI has fallen faster than any other transformative technology in history!"

6. Overlooked Insights

The comparison set implies AI's trajectory resembles a steep sequencing-style curve, not a gradual infrastructure curve

The chart includes DNA sequencing, a technology whose cost collapse (2001–2022) enabled entire new industries, as the closest analog to AI's curve. This hints that AI cost declines could unlock new categories of demand, not just cheaper versions of existing use cases.

  • Chart labels: "DNA sequencing 2001–2022" plotted alongside "AI 2021–Today"

AI's line is still short, so extrapolation carries uncertainty

AI's curve covers only about 5 years of data versus 20 to 80 years for the others. Other technologies' declines flattened over time, and it's unknown whether AI's will.

  • Chart axis: "Years since the start of price decline" (AI's line ends near year 5, while Compute extends to roughly year 61 and Electricity to roughly year 81)