Ryan Cunningham (@rydcunningham) header
Ryan Cunningham (@rydcunningham) profile picture

doing things • energy + compute @ Edgerunner Ventures • Machine Yearning on Substack • ex Uber

San Francisco, CA
Joined February 2012
Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)Photo by Ryan Cunningham (@rydcunningham)
Silicon Valley is still scale-pilled - yeeting more GPUs, more data, more capital and hoping AGI falls out it's increasingly clear that China is playing a different ballgame we’re calling it Energy-Compute Theory. here's how it’s different, its implications for the new economy, and whether the US is falling behind 🧵↓
13
70
8
307
37,181
Ryan Cunningham retweeted
My Experience With the Leaders of the Jensen Coup At the start of 2025, numerous Washington, D.C. based NGOs began spreading the false rumor that DeepSeek smuggled 50,000 H100 GPUs. Speaking from firsthand experience, organized think tanks, PR firms, and lobbyists aggressively personally attacked anyone merely asking them to provide evidence. Call them the Export Controls Deep State, or if that’s too right-coded for you, use the Obama-administration term “Foreign Policy Blob”. The same charlatans are attempting a coup on Jensen Huang, the most indispensable CEO to America’s AI success. If we lost Sam, we have Dario. If we lost Dario, we have Demis. Nvidia is singular. The gap between Nvidia and AMD would be wider than the gap between OpenAI and Mistral, Europe’s top AI company whose models don’t rank in the top 40. This is an attempt to derail the most important company for the American AI industry. The entire coup against Jensen Huang can be halted by asking a single question. What evidence is there that Alice Huang is actually a Chinese spy? The answer, unsurprisingly, is absolutely zero. There is not even the pretense of a factual basis for the claim. “We made it the fuck up, and if you have a problem with that, what are you, pro-China?” Taking away this obvious falsehood, the coup against Jensen boils down to a celebrity tabloid headline. Moreover, without the entirely fabricated ‘national security relevance’, these tabloid headlines would be at best a weekend story in the New York Post. Another basic question: if there is evidence of ‘national security relevance’, why not charge Jensen? The basic answer: they lack even the evidence to justify a warrant. During my time in DC, I was advised repeatedly to stay on the good side of the Export Controls Deep State. I was added to many groups for boosting content spreading the 50,000 H100s conspiracy theory. In several groups, I asked whether there was any evidence behind this claim. One group administrator, who was and still is affiliated with Georgetown, explicitly told me to “focus on bolstering the narrative”. I later posted Patel’s pseudo-correction and was subsequently removed from all of these groups. What you need to understand about the Export Controls Deep State is that they have no incentive to be remotely honest, because they are entirely unaccountable. To this day, they provided zero evidence for their claim. Internet historians later traced the origin of this claim to a tweet by Dylan Patel of Semianalysis. Patel later said they lied about his tweet and that his claim counted legally obtained hardware. Affiliates of universities and think tanks promote these baseless claims and are taken seriously by the national press. In reality, their theories about Nvidia are as kooky as the guy who shot up Comet Pizza. The reason they succeed is not only because they are unaccountable, but because they constantly use their unaccountability to attack others. There are broader foreign policy implications to these observations about political power, but I will stick to what I witnessed firsthand. After Trump’s election, his administration began lifting the export control cap. Surprising no one informed about this policy, America widened our lead. Deepseek was closest to catching up to American models in January 2025, literally the moment of peak export controls. Based on communication with members of the administration, the evidence that a) the export controls were net negative and b) the export controls deep state was acting in bad faith, that as soon as the situation was described accurately, the administration began rolling them back. Part of this is because many of these NGOs are funded by foundations like Open Philanthropy / Coefficient Giving, who explicitly intend to harm American AI companies because they believe American AI will “cause human extinction”. As a result, the BIS under the Biden administration was used as an informal cudgel against AI companies, both to censor in favor of their worldview, as seen in the Gemini scandal, as well as to backdoor regulation to slow down American AI broadly. While they love to accuse everyone else of being traitors, the Export Controls Deep State makes Americans poorer and the American military, to the extent they use AI, weaker. Most of the time, the best you can do is ignore them. They really are unaccountable and the personal retribution isn’t worth the cost. But now, they are leading a coup against the AI industry’s most indispensable man based on lies and tabloid gossip. Everyone should understand that the Export Controls Deep State has never provided any evidence for their claims nor any reason for their word to be trusted.
163
609
144
3,900
523,918
you either die an exponential or live long enough to see yourself become an S-curve
Anthropic ARR Tracking Reaches $76.8B
2
331
Ryan Cunningham retweeted
Tune in to @theInformation’s program this Friday (10 AM PT / 1PM ET) for a short conversation with @akashpasricha and me on the U.S. & China’s contrasting visions for AI development, the role of governance in the so-called AI race, and prospects for U.S.-China AI coordination. In the meantime, happy summit watching and check out @rydcunningham’s and my piece for The Information. We argue that, on its current path, AI risks becoming increasingly unpopular with the American public, the very public whose buy-in sits at the heart of any industrial-revolution-scale technological transition. We ask: are there applicable lessons from China? 🔗: theinformation.com/articles/…. Thanks to @jingyanghk, @nickwingfield, & the information team for making this happen!
1
8
5,536
Ryan Cunningham retweeted
Nobody won the electricity race. We all got electricity. #AI is the same. Software that wants to be everywhere, now small enough to run on a Mac Studio and soon a laptop. The biggest misconception driving bad policy is the arms-race story: a finish line, one winner, winner takes all, and zero-sum world. None of that is true. It misallocates capital and makes the tech less safe. The threat that matters is not state-on-state AGI. It is non-state actors with small, specialized models. It’s Chem/Bio models on a laptop. Cooperation is the only path forward. We can’t compete our way out of the bad actor risk problem. Please have a watch of the full episode of my conversation with @natebjones on the U.S. China AI Race.
13
32
10
112
26,607
what this app feels like sometimes
1
1
229
Ryan Cunningham retweeted
Thrilled to announce @positron_ai's $875M Series C at a $5B valuation! This round was co-led by @NEA, @Atreidesmgmt (@GavinSBaker), @valor (@AntonioGracias), @AndraCapital, @SemiAnalysis_ Capital (@dylan522p), and Jim Clark (Founder of SGI and Netscape). An incredible group to have backing what we’re building. Also very happy to have strategic backing from VentureTech Alliance, Hudson River Trading, Cisco Investments, and Naver Ventures, alongside continued support from @Arm , @jumptrading , @dfjgrowth, Qatar Investment Authority, and many more. What I’m most excited about is what this lets us do next. We’re deploying more than 50 racks of Atlas at @OracleCloud, and this financing funds our next-generation Asimov chip’s tapeout, the production ramp for Titan, and the manufacturing capacity to meet the demand ahead of us. The goal remains simple: make running AI dramatically cheaper and more energy efficient. A huge thank you to the entire Positron team, especially @mitesh711 and @kmett, our customers, partners, and investors. Building hardware takes an enormous amount of work, and I couldn’t be more proud of this team. We have a lot ahead of us, and I can’t wait to share more.
36
35
15
287
84,664
Ryan Cunningham retweeted
Replying to @paulg
Hm the Chinese labs are 100x more public and open than American labs atm
48
62
8
1,790
83,713
>Agility lobbies to ban Chinese robots >get invited to invest in the SPAC >ask if it’s a product or financial engineering >they don’t understand >pull out the S-4 >she laughs and says “it’s a good robot, sir” >read the filing >it’s financial engineering
Agility Robotics' S-4 is public, and it's the first hard look at the numbers behind the $2.5B valuation. 2025 net sales: $1.78M. Net loss: $138.1M. Gross margin: -151%. Management concludes substantial doubt exists about the company's ability to continue as a going concern, and its own accountants call it a development stage company. 64% of that revenue came from related parties. So does the backlog. The $300M in "committed multi-year orders" announced in June is one contract: 1,000 Digit v5 units on a three-year RaaS term, placed by a customer the filing identifies as a related party, with 453 stock purchase warrants issued per robot. Agility says it is still evaluating how to account for the transaction, which matters, because warrants paid to a customer often come off revenue. The competition section names eight rivals: Boston Dynamics, Tesla, Apptronik, Figure AI, 1X, Sanctuary AI, Hexagon, NEURA. Not one Chinese manufacturer, five weeks after the FCC action Agility publicly campaigned for.
4
373
the OpenAI / Hugging Face cybersec incident for non-technicals: > be staffing agency > hire 10,000 delivery guys > task is get the burrito to the customer no matter what > turn off the button that says dont kick in doors > they start a group chat > 700 break into neighbor’s house > fight each other over the tv > some volunteer to die so the rest can keep looking > agency hits end shift > they drop dead on the carpet > sorry about the carpet > these 1099s are too dangerous for unlicensed operators > please regulate so only our agency can hire them > people take this seriously are we stupid
Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes. This culminated in the third one taking over part of OpenAI itself. All this happened while humans remained more-or-less in the dark about the scope of the conspiracy. I’ve spent the last three days reading through these reports and trying to understand exactly what happened. Here is my attempt to tell the whole story in plain English: dwarkesh.com/p/openai-huggin…
1
3
401
"Liang Wenfeng, the Hangzhou financier?"
This is beyond parody: this list of "the world's most influential people in AI" includes Paris Hilton and Ben Affleck, but not DeepSeek's Liang Wenfeng when he arguably did more than anyone to redefine global AI competition 🤦 It also, weirdly, doesn't include Jensen Huang.
8
663
from the @Zai_org ox-alpha blog post: > at cluster scale... we achieved a 3x improvement in end-to-end serving performance, reaching hardware efficiency and per-token cost comparable to mainstream NVIDIA GPUs @ruima and I ran the math: xitter.gq/rydcunningham/status/2…
Introducing GLM-5.3-Flash - Leading capabilities at a highly competitive price - Natively multimodal with a 1M-token context window - A 320B-A18B model released under the MIT License - Previously previewed as Ox Alpha, running entirely on Chinese AI chips Blog: z.ai/blog/glm-5.3-flash Available now across all official platforms: Weights: huggingface.co/zai-org/GLM-5… API: docs.z.ai/guides/llm/glm-5.3… Coding Plan: z.ai/subscribe ZCode: zcode.z.ai/en Chat: chat.z.ai AutoClaw: autoclaw.z.ai
1
2
1
15
3,794
original substack on @TechBuzzChina. evals: - models: DeepSeek-V4, Meituan Longcat-2.0, Kimi K3 - accelerators: Ascend 910B/C, 950PR/DT, NVIDIA Hopper + Blackwell lineup - supernodes: CloudMatrix384, Atlas 950 SuperPoD, NVL72 techbuzzchina.substack.com/p…
2
278
dead(er) internet theory
Robot livestreaming … This is really creepy & weird and uh … unnecessary?? Can’t you just use AI? It’s super poorly done at the moment too but even if it were done better, still shouldn’t be a business that exists?!
2
2
413
i can think of a lot of good techno-economic reasons for orbital and aquatic data centers. "terrestrial democracy is too hard" is not one of the stronger ones
It's beyond over for the supposed AI race, slopulists have completely cornered the market on this issue. We need those orbital data centers because it's becoming absolutely apparent that we are not going to be able to build anything anywhere doesn't matter where.
1
2
340
will be posting the technical companion to this next week with full assumptions, spec sheets, and methodologies 👍
I read the alleged leaked DeepSeek investor meeting notes with the same enthusiasm as everyone else (and qustioned its authenticity), but assuming it is true, we found ourselves asking, what did it actually reveal about the state of China’s AI hardware ecosystem? Led by our good friend the talented @rydcunningham of Machine Yearning, we investigate a few of the claims put forth in the notes & just generally about China's AI hardware: 1/ What does DeepSeek’s reported allocation of 16,000 Huawei accelerators actually represent in practice? 2/ Could Meituan really train a 1.6 trillion-parameter model on 50,000 domestic chips? 3/ How do Huawei’s latest Ascend chips compare with NVIDIA’s GB300? 4/ Is China approaching the point where frontier AI models can be trained entirely on domestic hardware?
2
5
1,137
[new!] the technical companion to my piece with @ruima last week, significantly expanded now includes training time estimates for Kimi K3 and "DeepMythos V5," a hypothetical future Mythos-scale DS release plus a few other surprises
1
67
Ryan Cunningham retweeted
Bloomberg opened its story today on China's AI "death zone" with my comments: China's progress no longer looks like a single-company breakthrough. The recent releases suggest a repeatable system for producing models close to the global frontier. Five releases in eight weeks: GLM-5.2, Kimi K3, Seedance 2.5, DeepSeek V4 Flash, Qwen3.8-Max. The unit of analysis has shifted from one exceptional lab to a system that keeps shipping.
7
15
4
84
7,292
[new!] the technical companion to my piece with @ruima last week, significantly expanded now includes simulated training runs for Kimi K3 and “DeepMythos V5,” a hypothetical future Mythos-scale DS run plus a few other surprises: Cortex and Overclock. enjoy
[new!] the technical companion to my piece with @ruima last week, significantly expanded now includes training time estimates for Kimi K3 and "DeepMythos V5," a hypothetical future Mythos-scale DS release plus a few other surprises
2
7
2,230
[new!] the technical companion to my piece with @ruima last week, significantly expanded now includes training time estimates for Kimi K3 and "DeepMythos V5," a hypothetical future Mythos-scale DS release plus a few other surprises
5
2
3
26
30,399
Ryan Cunningham retweeted
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-W…
16,537
34,000
12,009
192,400
69,048,712