Is Deepseek more an attack on Wall Street than an AI innovation?
The launch of Deepseek has been framed as the AI race’s Sputnik but the timing is dubious and questions are being asked about whether the tech is really as game-changing as its creators claim. Either way, America must stay vigilant, says Mark Minevich
It started with a single, seismic headline: China builds an AI model for just $6m.
The claim seemed almost too bold to be true. In an era where Western AI giants pour billions into training their latest large language models (LLMs), the number felt impossible – an engineered disruption, a bombshell lobbed directly at the heart of Silicon Valley’s AI dominance.
At the centre of the storm: Liang Wenfeng, the enigmatic founder of Deepseek, China’s rising AI powerhouse. A hedge fund veteran turned AI visionary, Wenfeng wasn’t just announcing a technical breakthrough – he was setting off a financial and geopolitical earthquake.
But behind Deepseek stands an even more intriguing player: High-Flyer (HF), a Chinese quant hedge fund that doesn’t just invest in AI but actively manipulates markets, plays financial war games and has a history of regulatory violations.
The market reaction was swift and brutal. Nvidia plunged nearly 17 per cent in a single day, erasing $593bn in market value – one of the largest single-day losses in stock market history. TSMC fell 14 per cent, dragging semiconductor and chip-related stocks down with it. AI infrastructure and tech stocks followed suit, with companies tied to cloud computing, data centres, and GPU supply chains experiencing heavy sell-offs.
A coordinated financial manoeuvre?
The shock wasn’t just about Deepseek’s AI model itself – it was about what it represented. Almost immediately, speculation erupted that this wasn’t just an AI breakthrough – it was a coordinated financial manoeuvre.
Multiple sources on social media, including X, claimed that High-Flyer used $600-800m to short US equities in the days leading up to Deepseek’s announcement. If true, this means they could have made $6-8bn in profits in just one day.
Deepseek’s announcement didn’t just shake tech stocks – it forced Wall Street to question the AI bubble itself. Can AI investment remain sustainable if China claims it can achieve similar results for a fraction of the cost? Are Nvidia’s massive valuations still justified? If AI development costs plummet, will current business models survive?
For months, investors had been pouring billions into AI stocks, assuming that AI’s inevitable progress would drive massive, long-term profits. But Deepseek’s claim forced the market to reassess its assumptions overnight.
Deepseek did what Chinese companies are known for: copying. One month ago, Deepseek had no world-class model. Now, suddenly, it does? The reality is that some copying has been done in one way or another. According to Donald Trump’s AI tsar, David Sacks, Deepseek has likely distilled ChatGPT-4, making it cheaper, but not necessarily better.
This doesn’t mean OpenAI, Google or Anthropic were foolish to invest billions of dollars. The cost of developing new cutting-edge models is enormous. If a company can copy the breakthroughs without the R&D costs, they can drastically reduce expenses.
That’s why Deepseek is cheap – it’s not innovation, it’s imitation.
There’s another Deepseek mystery: is it really that cheap? As Scale AI CEO Alexandr Wang pointed out, Deepseek reportedly has around 50,000 Nvidia H100 GPUs that they can’t openly discuss due to US export controls.
If true, the efficiency gains they’re claiming aren’t as impressive as they seem because DeepSeek isn’t achieving world-class AI on a shoestring budget. Instead, they may have simply acquired restricted GPUs through indirect means and used them to train their models just like OpenAI or Anthropic.
The path forward is to learn, adapt, and accelerate. President Trump, rescinding Biden’s executive orders that hamstrung American AI companies, has opened up the faucet for AI innovation, and put David Sacks at the head of American AI and crypto efforts.
As famous VC investor Marc Andreessen put it, “Deepseek R1 is AI’s Sputnik moment”. We must rise to meet it.
The AI Race Is just beginning. Deepseek is a warning shot – but not a knockout punch. America remains the leader in AI, but it must stay vigilant. If American companies absorb Deepseek’s optimisations while leveraging top-tier GPUs, we will propel past this moment even faster.
Welcome to the new AI frontier.
Mark Minevich is the founding partner of Going Global Ventures