Nathan Elmore

Journalist. Frontier Models: The Minds Behind AI.

[Nathan Elmore @NatElmore]

Liang Wenfeng, the mind behind DeepSeek, built a challenger reshaping the global AI race.
This book frames his rise as more than a startup story: it’s a clash of innovation, geopolitics, and the pressure of chip restrictions on China’s AI ambitions.
https://t.co/ScBUkQEaWa

[Nathan Elmore @NatElmore]

@pbillore141 Read the full story:
https://t.co/1GC7Zh9x3i

[Nathan Elmore @NatElmore]

Full context: https://t.co/F4l6nej62c

[Nathan Elmore @NatElmore]

DeepSeek's founder Liang Wenfeng grew his wealth by 3,850% to top the global billionaire growth list — one of the fastest wealth surges ever recorded for a tech founder.

[Nathan Elmore @NatElmore]

Most Western AI founders got rich on paper through valuations. Liang's gain reflects actual market recognition of DeepSeek as a serious competitor to OpenAI and Google.

[Eren Chen @ErenChenAI]

DeepSeek founder Liang Wenfeng is now the world’s richest self-made AI entrepreneur, with an estimated net worth of $39.5 billion — ahead of the founders of Anthropic, Scale AI, and other major AI startups.

Meanwhile, DeepSeek has become the #1 model provider by token usage on OpenRouter, capturing 18% of all tokens on the platform due to its low cost.

And DeepSeek Harness is coming soon.

[Nathan Elmore @NatElmore]

More context on the broader industry shift: https://t.co/hXYfSWcxpm

[Nathan Elmore @NatElmore]

DeepSeek's API was priced so low it likely ran at a negative gross profit margin — and now a "significant increase" is coming. The era of floor-price AI may be over.

[Nathan Elmore @NatElmore]

A 3x hike would push input from 1 yuan to 4 yuan per million Tokens and likely flip gross margin from negative to 30–40%. A 4x hike still lands below what domestic competitors charge — but shrinks the price gap from 10x to roughly 2x.

[Zephyr @zephyr_z9]

Email from the Whale

[morgan — @morqon]

deepseek’s liang wenfeng on raising prices, from the leaked investor call

“if we wanted to maximise profit, we would charge more. at this price point, demand is inelastic: if i raised the price by another 50 per cent, or even doubled it, token consumption would not change very much. if i doubled the price, total revenue would almost double.”

[meowmeow @meowmeownmtb]

@teortaxesTex 梁文锋大学毕业的时候 这是他小学同学曾经在网上发的

[Nathan Elmore @NatElmore]

@teortaxesTex @jd_pressman @giffmana There’s very little information about Wenfeng, only interviews. I don’t think it’s deliberate; it may just be a mix of low awareness of what he’s doing and a lack of personal vanity.

[Nathan Elmore @NatElmore]

@teortaxesTex Wenfeng achieved his mastery in AI by developing stock trading algorithms at High-Flyer.

[Nathan Elmore @NatElmore]

@teortaxesTex It’s hard to make any claim about autism, but in the journalistic work I did for my book, I also detected that obsessive pattern, which can be seen in many highly intelligent people.

[Nathan Elmore @NatElmore]

Very interesting personal details and photos about Liang Wenfeng. Recommended:

[动察Beating @BeatingOfficial]

https://t.co/aRL5putzx2

[Nathan Elmore @NatElmore]

Full details and sourcing at https://t.co/dQjd73UL1k

[Nathan Elmore @NatElmore]

DeepSeek's new V4-Flash model is reportedly 100x cheaper to run than Anthropic's Claude, according to a research firm; that's not a rounding error, that's a structural cost collapse.

[Nathan Elmore @NatElmore]

DeepSeek is a Chinese startup, not a hyperscaler. Matching or undercutting frontier US labs on price, while staying competitive on capability, puts real pressure on Anthropic's and OpenAI's pricing models.