- That's so Meta
- Posts
- Special Edition: Meta vs China AI
Special Edition: Meta vs China AI
Analysis of Chinese AI infra and what it's like to compete with a nation state

Ahoy Metamates! š¢ š“āā ļø
BIG TLDR for those of you freaking out about DeepSeek. DeepSeek just handed Meta the entire AI market. Congratu-f*cking-lations, fellow shareholders. Skip to the 3rd section if youāre unable to contain yourself.
But first, a story on how AI became a sudden emergency:
Hereās a scary thought: when the FBI was looking for a Head of AI last year, they interviewed my co-founder. Now, David is wicked smart. But I think heād be the first person to tell you we submitted his resume there at 3 AM after a dare with some DEFCON friendz and a respectable amount of scotch. He goes to the interview and they tell him the job for Head of AI of the FBI has actually nothing to do with AI but itās all about writing policies that define what kinds of AI should the FBI take seriously.
Thatās all to say: weāre f*cked.
I try to be polite about these things but, weāre f*cked.
Being on the cutting edge of AI is not just a feel good pride thing for us tech folk, itās a national security capability for the US both in economic and military warfare. Non- linear technological progress is a highly volatile time where the balance of power in the world can shift rapidly, and possibly irreversibly. In this case, it can shift in Metaās favor.
While most people were highly obsessed with Mark Z staring at Lauren Sanchezā cleavage, I found it pretty telling that he was in the front row at the inauguration to begin with. In case I need to be explicit: weāre basically in a cold war with China and Meta + Palantir + Whatever it is that Elon does = Bell Labs.
Hold on to your britches because most of your favorite logos will struggle on the stock market this year.
First-
Some background from Marc Andreessen (a key Meta board member):
When this video came out, the immediate public conclusion was that Marc is red pilled, likely pumping T and probably spending a lot of time on 4chan. However what I hear is:
The government is using last ditch effort tactics of direct interference and comparing that to āphysics in the Cold Warā
Logical conclusion is they consider AI to be of critical offensive capability
This level of desperation means weāre in deep trouble
Relevant to us because Marc is on Metaās board + Meta has made AI a priority and has been blocked at every turn by the government on infra + Meta vs China is the most relevant tech battle of our times, other than semiconductors.
Next, some background on Meta AI efforts
Meta has the best strategy on AI in the US - commoditize their models to ensure they get wide adoption and eventually drive down costs through efficiency of scale. It was also working much better than most people realized (up until last week).
ChatGPT, despite its popularity, only had 5% of Fortune 500 as clients
Llama enterprise has real heavy hitter clients like Block, Spotify, LinkedIn, IBM etc.
Best explanation Iāve seen of why is this video ( go to timestamp 1:05:46) but TLDR: faster adoption leads to better lock-in with ads.
On getting PWNED by China - Meta wins, actually
With all this said - The degree to which weāre f*cked by China is currently NOT as bad as people are tweeting about ( Iād say weāre medium f*cked). Bares pointing out, the single person who saw Chinese AI coming around the corner is Mark Zuckerberg, who from the beginning made the risky bet that AI will commoditize. Metaās strategy with Llama is pretty genius.
TLDR on what the actual deal is with Deep Seek - click to read whole thread. Itās currently the only non-bullsh*t tech boi take on this Iāve seen:
Okay. Thanks for the nerd snipe guys. I spent the day learning exactly how DeepSeek trained at 1/30 the price, instead of working on my pitch deck. The tl;dr to everything, according to their papers:
ā wordgrammer (@wordgrammer)
3:04 AM ⢠Jan 27, 2025
But basically - here goes the controversial take:
DeepSeek made 2 major cost cutting innovations
This makes them much more efficient at scale
Export restrictions are still in place (Chinaās ability to get more NVdIA H100s is limited)
Meta is sitting on 600,000 H100s
Thanks, China for making AI more efficient
Meta will crush the market now
OK thanks #bye
Provided that Brian Nowak at Morgan Stanley doesnāt try to sink the stock with his lack of understanding about technology AGAIN, we should have a linear path to Meta becoming one of the most valuable businesses ever built.
Understanding Chinese Tech
Most people have a negative view of Chinese tech as āinferiorā - and thatās a mistake. Beyond the 996 work culture and the relentless attempt to steal patented tech, thereās a drive thatās nearly forgotten in Silicon Valley: the need to prove youāre best.
Thatās right. While youāre TC or GTFO-ing on Blind, your salary alone is likely more than what DeepSeek cost to build in its entirety. Competing against a hungry person is a dangerous gambit. And most of us tend to forget where we came from when the bank account hits 7 figures.
Not to be harsh but yāall got murdered by simply being lazy.
OK, getting off my soap box - hereās the data:
China has been aggressively and directly funding AI work for 20+ years

Among the top 155 AI projects:
54 (35 percent) went to the health sector,
39 were government and civil society projects,
and another 38 were allocated to communication technologies
On average, the 30 health-sector projects cost $11 million per project.(RAND data). Iām not even going to pull data to compare that with how much funding US AI startups get because it would be comical.

And while weāre trying to build widgets to impress VCs here in the US, Chinaās AI tech for the last decade has been a tool for diplomacy wins:

Most of it supported some form of dictatorial / oppressive political regime. Maybe if youāre feeling extracurricular today, Google why Congress investigated US VC Firms Sequoia and GGV investing in Chinese AI that was used in active combat zones in some cases.

Now, About Military Capabilities
The other weird thing about Communism is that priorities are communicated clearly - unlike US politics, what is said is actually meant.
Here are Chinaās explicit tech priorities:

Source: RAND Research 2024
Now, itās not a secret to anyone that the Peopleās Liberation Army is using AI in a myriad of ways, including offensively to wage infowar on adversaries (thatās us).
Historically, the PLAās influence operations generally fell under the concept of the āThree Warfaresā (äøē§ęę³ ), which were developed in the mid-2000s and included āpsychological warfareā (åæēę), āpublic opinion warfareā (č论ę), and ālegal warfareā (ę³å¾ę). Cyber-enabled influence operations generally fell under public opinion warfare, specifically āonline public opinion warfareā (ē½ē»č论ę).
But it may come as a surprise that they feel this is a necessary defence tactic:
Much of the surveyed PLA writings are focused on the potential cyber and AI-driven information threat from the United States. For example, there is a lot of euphemistic discussion about concerns for CCP regime security stemming from AI technology, ranging from the fact that Western large language models (LLMs) are inherently biased toward Western values to the risks of the United States and other Western countries using generative AI against the CCP.
And being that this is existential, they didnāt waste time operationalizing it:

Long Yameng [é¾äŗč] and Zhou Yang [åØę“], āResearch on the Application of Social Bots in Public Opinion Struggleā [ā社交ęŗåØäŗŗåØč 论ęäŗäøēåŗēØē ē©¶ā],
I know this is a lot of words- so letās do a palate cleanser picture.

TLDR: AI is an offensive military capability and developing DeepSeek was existential because of fear US based models are biased and can influence society.
Whatās Next
Earnings call in two days!
Meta might buy TikTok ( but probably not) and for some reason Perplexity is bidding on it? Make it make sense
Iāll be back with analyst reports next week
The information provided herein is for general informational purposes only and is not intended to provide tax, legal, or investment advice and should not be construed as an offer to sell, a solicitation of an offer to buy, or a recommendation of any security by That's So Meta, its employees and affiliates, or any third-party. Any expressions of opinion or assumptions are for illustrative purposes only and are subject to change without notice. Past performance is not a guarantee of future results and the opinions presented herein should not be viewed as an indicator of future performance. Investing in securities involves risk. Loss of principal is possible.Third-party data has been obtained from sources we believe to be reliable; however, its accuracy, completeness, or reliability cannot be guaranteed.