Navigating the New AI Policy Landscape: A Call for Balance - Independent Review

Why the New U.S. AI Policy Has Me Concerned—and Why You Should Be Too

The U.S. government just released a new AI policy this morning titled "Ensuring U.S. Security and Economic Strength in the Age of Artificial Intelligence." As someone deeply invested in the future of AI, I feel compelled to share why this policy, despite its good intentions, may end up doing more harm than good—especially for the innovation and collaboration that have been driving AI forward.

This isn’t just an AI industry problem. It’s a you and me problem because the decisions being made today will shape the way AI impacts all our lives—from healthcare to education to the jobs we hold tomorrow.

Let me break it down.

A Policy That Could Hold Us Back

The policy creates new rules for how advanced AI tools (like powerful graphics processing units, or GPUs) are shared internationally. For countries that aren’t considered close U.S. allies, it caps access to these tools at 50,000 GPUs per year. On paper, this might sound like a smart way to keep AI technology safe, but here’s the catch: AI isn’t just about who has the tools. It’s about how we use them to innovate and solve problems.

By imposing these restrictions, we could unintentionally shut out smaller innovators—startups and researchers without deep pockets or global networks—who are often the ones driving breakthroughs. Think about it: Would we have had something like ChatGPT if a small team of researchers couldn’t access the resources they needed to experiment?

This policy could also make other countries step up their game, offering more open and flexible environments for innovation. If that happens, the U.S. risks losing its edge in AI—a space where being a leader is everything.

Lets compare the 50,000 GPU's per year to the U.S. and Global Market in 2023 and early 2024:

• Nvidia's Data Center GPUs: Nvidia reported shipping around 3.76 million data center GPUs in 2023.
• AMD's Desktop GPUs: While exact annual figures for AMD's GPU sales in 2023 weren't provided, one document mentions that AMD shipped 1.09 million desktop AIBs (Graphics Add-in Boards) in Q2 2023 alone. Given the growth in their market share and sales throughout the year, including other quarters, we can estimate a significant but unspecified number for the entire year.
• Total GPU Market: The total market for GPUs, including both discrete and integrated, saw shipments of 70 million units in Q1 2024 alone, with a year-over-year increase of 28%. This suggests that 2023 was a year of significant recovery and growth in GPU sales.

Given these points, here's a rough estimation:

• Nvidia: 3.76 million data center GPUs (excluding consumer GPUs).
• AMD: Considering their market share and growth, let's assume at least 4-5 million desktop GPUs for the full year, although this is a conservative estimate based on one quarter's data.
• Intel: While specific numbers for discrete GPU sales aren't given, Intel holds a significant share in integrated GPUs, which are included in CPU shipments. For discrete GPUs, let's estimate a lower number, perhaps around 1 million (given their newer entry into this market).

Total Estimate for 2023:

• Adding these together, we get:

This gives us a total of roughly 9.26 million discrete GPUs from major U.S. companies for data center and desktop use. However, this does not include:

• Integrated GPUs: Intel, being a major player in integrated graphics, likely contributes to tens of millions more when including CPUs with integrated graphics.
• Total Market: Since we know the total GPU shipments for Q1 2024 were 70 million, and considering this was a period of growth, we can infer that the 2023 total market might have been in the range of 50 to 60 million GPUs for all types, including integrated, desktop, and data center GPUs from all manufacturers worldwide.

Thus, the best estimate for the total number of GPUs sold in 2023, including all types, would be around 50 to 60 million units globally. This number encompasses U.S. companies' contributions along with other global players.

Why Collaboration Matters More Than Ever

AI doesn’t exist in a vacuum. It’s built on collaboration—across borders, industries, and cultures. The policy’s emphasis on classifying countries into “allies” and “concerns” risks fragmenting the global AI community.

Here’s why that’s a problem: AI has the potential to solve issues that affect everyone, like climate change or global health crises. But if countries are operating in silos, we lose the ability to create shared ethical standards, safety measures, or solutions that benefit everyone.

Instead of drawing hard lines, the U.S. should be bringing others to the table, leading conversations on how we can all use AI responsibly and effectively.

The Unintended Consequences of Too Much Control

Now, let’s get to the part that really concerns me: the impact on small businesses and the wider AI ecosystem.

The policy creates new hoops to jump through, like licensing requirements for GPU orders that go beyond a certain size (1,700 GPUs, to be exact). This might not sound like a big deal, but for startups and small companies, extra bureaucracy means less time and money for innovation.

Here’s an example: Meta is planning to use 350,000 GPUs for their AI training efforts. If the limit for others is 1,700, you can see how the playing field isn’t just uneven—it’s practically tilted.

These limits could force small companies to slow down or even give up on projects, meaning fewer breakthroughs and less competition to push the whole field forward.

The Cake Analogy: Why Restricting AI May Backfire

To explain why this policy might not have the effect we want, let’s talk about cake. Yes, cake.

Imagine you’ve created the world’s best cake recipe (think of the "model weights" which are essentially the ingredients and steps to make AI work).. It’s so good that you decide to lock it away to keep it safe. Only a few trusted friends get to bake it. Sounds smart, right?

But here’s the problem: By keeping the recipe under lock and key, you push others to come up with their own recipes. Maybe they won’t follow your safety standards. Maybe their cakes will taste worse—or even be dangerous. Either way, you’re no longer leading the cake game.

The same thing happens with AI. If the U.S. restricts access to key resources and technology, we’re not stopping others from building their own systems. We’re just giving up the chance to influence how they’re made—and that’s where real security comes from.

Instead of hiding our “recipe,” we should be setting the standard for how AI is developed and used—ethically, responsibly, and with the future in mind.

What This Means for You

AI isn’t just a tech industry buzzword anymore. It’s already changing how we work, learn, and live. This policy could affect all of us in ways that aren’t immediately obvious:

• Fewer breakthroughs in healthcare technology that could save lives.
• Slower adoption of AI tools that make businesses more efficient.
• A less competitive U.S. economy, meaning fewer opportunities for American workers.

If we want to stay at the forefront of AI, we need policies that empower innovation while keeping us safe. That’s not an easy balance to strike, but it’s worth striving for.

What Needs to Change

1. Support Small Innovators: Make it easier for startups and researchers to access the resources they need, without unnecessary red tape.
2. Lead Through Collaboration: Bring global players to the table to create shared ethical standards for AI.
3. Focus on Agility: Regularly update policies to keep up with AI’s rapid evolution, rather than locking us into outdated frameworks.
4. Champion Openness: Set the global standard for safe, ethical AI development, instead of shutting others out.

A Call to Action

Policies like this one don’t just affect tech companies—they shape the future we’re all going to live in. It’s up to all of us to demand a smarter approach—one that protects our security while fostering innovation, collaboration, and progress.

Let’s make sure we’re building an AI ecosystem that works for everyone, not just the few who already have the keys to the kingdom.

I hope this breakdown helps :)

~JP