- ALFA Institute
- Posts
- DeepSeek Reality Check: China's Good Model, But America Can Still Win Big
DeepSeek Reality Check: China's Good Model, But America Can Still Win Big
Are you paying attention yet?
Monday’s market moves sent shockwaves across the country. The past, present, and future implications of the DeepSeek developments all played out in real time across news screens and X feeds everywhere.
It’s hard to mistake the historical parallels in this week’s unexpected technological jolt. Andreessen Horowitz’s Alex Rampell details nearly seventy years ago:
On October 4, 1957, the USSR launched a 184 pound satellite into orbit: Sputnik.
The satellite didn’t do anything — it just “beeped” (over radio waves). But it was an alarm clock for the USA.
Shortly after Sputnik 1 (and Sputnik 2 in November 1957), the US attempted to respond with the Navy’s Project Vanguard on December 6, 1957, which exploded 2 seconds after liftoff.
Almost immediately, Eisenhower greenlit the Army’s Jupiter-C/Explorer program under Wernher von Braun, which launched Explorer 1 — America’s first successful orbiting satellite — in January of 1958.
Shortly thereafter was a flurry of activity: Congress and Eisenhower established NASA in 1958, consolidating military and civilian efforts into a unified agency. NASA’s funding grew 60x from 1958-1964.
Congress passed the National Defense Education Act that same year, channeling $1B into STEM education, and marking a cultural shift, encouraging young Americans to pursue careers in engineering.
ARPA (now DARPA) was created in February, 1958 (the Internet comes from ARPANET!).
America went from laggard in the space race to unified, fast-moving behemoth. By 1969, we had landed men on the moon, a feat the Soviets never accomplished. Sputnik didn’t just wake us up to pursue and win the space race, but arguably renewed America’s commitment to scientific leadership and engineering prowess, laying the groundwork for America’s role as a global technology leader in the decades to come.
While the analogy isn’t perfect, DeepSeek should also serve as a blaring alarm to those who thought that America’s lead in the AI race was inevitable or insurmountable.
Yet this moment of "catch-up" comes with a crucial asterisk – it was achieved using deliberately handicapped chips under a previous export control regime. The question isn't whether China can compete; they've proven they can.
DeepSeek’s innovations in model architecture and training efficiency showcase the kind of relentless pursuit we should expect from a serious competitor.
Now, America is faced with the opportunity to make full use of its considerable advantages in securing its position as a leader in AI and global technological development.
Unleash American Energy Infrastructure
China is building enough power plants to drive their AI ambitions for the next decade with two new coal power plants every week and almost as many new nuclear power plants as the rest of the world combined
China’s accelerating electricity generation
America has better chip design and better research, yet we have been bottlenecking our advantage through insufficient energy infrastructure.
Over the past four years, the Biden administration pushed to decarbonize our economy and emphasized federal design over private enterprise. Every recent AI data center announcement has come with hand-wringing about power constraints.
We need to build more energy infrastructure. Immediately. President Trump’s announced energy emergency declaration will help with this. No more bureaucratic delays or frivolous lawsuits. We need to produce more energy, approve pipelines, fast-track licenses for nuclear projects, and upgrade our grid. It’s simple: no American power, no American AI dominance.
America's Moment
China just proved they can compete at a high level in AI on limited chips and compute. This isn’t a cause for panic – it's cause for getting serious about winning.
The fundamentals still heavily favor America: we have the best semiconductor design, the best research institutions, the most dynamic tech companies, the deepest capital markets, and incredible natural resources. But these advantages only matter if we use them. Every power plant not built, every allied coordination opportunity missed, every moment of complacency is a gift to our competitors.
DeepSeek isn't a warning about American weakness; it's a reminder about Chinese determination. They're playing to win. A few decades of watching Chinese companies mainly copy and commercialize Western innovations led many to dismiss their capacity for breakthrough work. That was a mistake.
The good news? Nothing about DeepSeek's achievements changes America's massive structural advantages. We're still the heavyweight champion – but we just got a clear signal that our challenger is training seriously. Time to stop coasting and start extending our lead.
The Trump administration has outlined an ambitious vision for American AI leadership through projects like Stargate. But the next true test will be execution: can we build the power plants, maintain allied coordination, and drive the kind of focused innovation needed to stay decisively ahead? The answer is simple: We must.
The next few years will determine whether the story of the 2020s is the American extension of technological dominance or a much messier competition.
Thanks for reading and have a great day!
🇺🇸🇺🇸🇺🇸