\n<\/p>\n
Andreessen Horowitz general partner and Mistral board member Anjney \u201cAnj\u201d Midha first spied DeepSeek\u2019s jaw-dropping performance<\/a> six months ago, he tells TechCrunch.<\/p>\n That\u2019s when DeepSeek introduced Coder V2, which rivaled OpenAI\u2019s GPT4-Turbo for coding-specific tasks, according to a paper<\/a> it released last year. This put DeepSeek on a path to release improved models every couple of months right through R1, he said. R1 is its new open source reasoning model that has upended the tech industry<\/a> for offering industry standard performance at a fraction of the cost.<\/p>\n Despite the sell-off of Nvidia\u2019s stock<\/a>, Midha says R1 doesn\u2019t mean that AI foundational models will stop spending billions to gobble GPU chips and build more data centers as fast as they can.\u00a0<\/p>\n It means they will do more with the compute power they can obtain.<\/p>\n \u201cWhen people are like, okay Anj, Mistral has raised a billion dollars,\u201d he says. \u201cDoes DeepSeek mean that all that billion dollars is completely unnecessary? No, actually, it\u2019s extraordinarily valuable for them to be able to look at DeepSeek\u2019s efficiency improvements, internalize them, and then throw a billion dollars at it.\u201d<\/p>\n He adds, \u201cNow we can get 10 times more output from the same compute.\u201d<\/p>\n That doesn\u2019t mean Mistral is hopelessly behind rivals OpenAI and Anthropic, he argues. Each of them have raised many more billions than Mistral. OpenAI is reportedly in talks to raise another jaw-dropping $40 billion.<\/a><\/p>\n Mistral remains competitive with them because it\u2019s open source, he says. And his logic does have merit. Open source gives a company access to essentially free technical labor from those who want to help because they use the project. Closed source rivals guard their secrets and have to pay for all the labor as well as compute power.<\/p>\n \u201cYou don\u2019t need $20 billion. You just need more compute than any other open source model app. So Mistral is positioned [well]. They have the most compute of any open source provider,\u201d Midha\u00a0said of his portfolio company.<\/p>\n Facebook\u2019s Llama, the largest Western open source AI model rival to Mistral, will also get plenty more investment. CEO Mark Zuckerberg on Wednesday said he\u2019s still planning to spend \u201chundreds of billions of dollars\u201d <\/a>overall on AI. That includes $60 billion in 2025 on capital expenditures, mostly data centers.\u00a0<\/p>\n Midha, who is also a board member for AI image generator Black Forest Labs and 3D model maker Luma<\/a> (and an angel in AI outfits Anthropic, ElevenLabs, and others) has another reason why he doesn\u2019t see AI\u2019s hunger for GPUs abating anytime soon.\u00a0<\/p>\n He\u2019s the leader of a16z\u2019s Oxygen program<\/a>. GPUs, particularly Nvidia\u2019s state-of-the-art H100s, have become such a scarce commodity that the VC firm took matters into its own hands about a year and a half ago. It bought a bunch of them for its portfolio companies to use.<\/p>\n Oxygen is \u201coverbooked right now. I can\u2019t allocate enough,\u201d Midha laughs. Not only do his startups need GPUs for AI model training, but then they need even more to run their ongoing AI products for customers.<\/p>\n \u201cNow there\u2019s this insatiable demand for inference, for the consumption,\u201d he explains.<\/p>\n That\u2019s also why he thinks DeepSeek\u2019s engineering breakthroughs won\u2019t change Stargate, either. That\u2019s OpenAI\u2019s big $500 billion partnership<\/a> announced earlier this month with SoftBank and Oracle for AI data centers.\u00a0<\/p>\n The major change DeepSeek ushers in is recognition by nation states that AI is the next foundational infrastructure, like electricity and the internet. Midha wants them to consider \u201cinfrastructure independence,\u201d as he calls it. Do they want to rely on Chinese models, with its censorship and claws in their data? Or do they want Western models that follow Western laws and ethics and abide by NATO agreements?\u00a0<\/p>\n He\u2019s obviously advocating for Western nations using Western models, like his Paris-based Mistral. Hundreds of companies share that concern and have already blocked DeepSeek,<\/a> which is both a consumer app service and an open source model.<\/p>\n Not everyone buys into that fear of Chinese open source models. Companies can run them locally in their own data centers. And DeepSeek is already available as a secure cloud service from American companies like Microsoft Azure Foundry<\/a>, so developers don\u2019t have to use DeepSeek\u2019s cloud service.<\/p>\n In fact, Intel\u2019s former CEO, Pat Gelsinger \u2014 someone well familiar with China<\/a> \u2014 told TechCrunch that his startup Gloo, is building AI chat services on their own version of DeepSeek R1<\/a> instead of choices like Llama or OpenAI.<\/p>\n But if anyone wants to ditch their data center plans in light of DeepSeek, Midra laughs and has a request: \u201cIf you have extra GPUs, please send them to Anj.\u201d<\/p>\n TechCrunch has an AI-focused newsletter!\u00a0Sign up here<\/a>\u00a0to get it in your inbox every Wednesday.<\/em><\/p>\n<\/div>\na16z\u2019s Oxygen GPU sharing program \u201coverbooked\u201d<\/h2>\n