> ## Content Index > Fetch the complete content index at: https://barbarosozturk.com/llms.txt > Use this file to discover other available public pages before exploring further. # We Won't Grow Our Way Out of $40 Trillion In Debt With Better Models. We Need Better Adoption. - URL: https://barbarosozturk.com/blog/we-wont-grow-our-way-out-of-40-trillion-in-debt-with-better-models-we-need-better-adoption/ - Published: 2026-08-30T15:32:43.000Z - Updated: 2026-08-30T15:32:43.000Z - Description: Washington thinks AI will let us grow our way out of $40 trillion in debt. The math says otherwise: interest is now growing faster than GDP. And Silicon Valley is solving the wrong problem. - Author: Barbaros - Tags: Economy & Markets, Technology, Artificial Intelligence There's a comforting story in DC right now: "Don't worry about the debt. AI will boost productivity, boost GDP, and we'll grow our way out of it, just like we did after WWII." I hear this in every VC pitch, every Bessent interview, every hyperscaler capex justification. The problem? The numbers no longer add up. And we're building the wrong thing to fix it. ![](https://storage.ghost.io/c/64/81/648165b7-979e-43dd-9c43-02101159f467/content/images/2026/08/2-4.jpg) ### The Debt Math Is Brutal With Washington claiming that the United States can “grow” out of its mounting fiscal burdens, some economists are sceptical that growth alone could absorb the country’s mountain of national debt, which just surpassed US$40 trillion. - The national debt crossed $39 trillion in March 2026, adding roughly $7.2 billion per day. Interest expense now exceeds $1 trillion annually — surpassing both the defense budget and Medicaid individually. - Net interest costs are projected to exceed $1 trillion in fiscal year 2026, amounting to about 3.3% of GDP and nearly 14% of federal spending this year. - Net interest payments will rise from 3.2% of GDP in 2025 to 4.6% in 2036, as the average nominal interest rate on government debt rises to exceed the nominal economic growth rate by 2031. Read that last line again: By 2031, the interest rate on government debt rises to exceed the nominal economic growth rate. One measure (interest payments as a percent of GDP) tells the tale. Unlike overall debt-to-GDP, interest payments show the flow of national income needed to service national debt. Between high deficits, growing debt, and higher interest rates, the share of GDP going to pay interest has doubled to around 3%. For the first time since the 1990s, we're in a world where borrowing costs are no longer anchored comfortably below the rate of economic growth. That's the textbook definition of an unsustainable debt path. CBO projects debt held by the public rising from 101% to 120% of GDP by 2036. ![](https://storage.ghost.io/c/64/81/648165b7-979e-43dd-9c43-02101159f467/content/images/2026/08/3-3.jpg) ### Yes, AI Could Help. But Not How We Think. Brookings ran the numbers: - In every scenario, AI improves the budget outlook, trimming projected federal debt by 39 to 49 percentage points of GDP three decades from now. - GDP growth assumptions matter most. Of all variables, the GDP growth rate has by far the largest effect on projected deficits, making AI's macroeconomic impact the key uncertainty. - This paper examines whether AI-driven productivity growth can resolve the United States' unsustainable fiscal trajectory, projected by CBO to push public debt to 175% of GDP by 2056 - A sustained 0.5 percentage point annual increase in TFP growth — if somehow achieved — would reduce debt held by the public by 12 percentage points of GDP. The optimistic case: Recent compelling independent economic research suggests that AI can raise productivity growth over the next 10 years by 0.9 percentage points from its current rate of about 1.8% to 2.7%, essentially a 3% GDP growth rate. That's huge. That's the difference between 2% and 3% GDP growth for a decade. But there's a catch: Absent a change in fiscal and healthcare policy, the deficit, currently around 6% of GDP, will continue to increase and even deepen. Higher productivity, while raising incomes, also raises interest rates and, therefore, federal interest. In other words: Even if AI works, higher growth = higher rates = higher interest payments. The debt treadmill speeds up. ![](https://storage.ghost.io/c/64/81/648165b7-979e-43dd-9c43-02101159f467/content/images/2026/08/4-3.jpg) ### We Are Solving The Wrong Bottleneck Look where all the money is going right now: $725B in hyperscaler capex, 90% into better models (GPT-5, Gemini 3, Claude 4) and better hardware (H100 -> B200 -> Rubin). Markets are obsessed with token cost, context windows, and FLOPS. But the real bottleneck is not compute. It's adoption. According to a recent report from Stanford, enterprise AI adoption has risen to 88%, up from 55% just two years ago. However, the vast majority of companies are still struggling to capture measurable economic gains: one report from the MIT Media Lab shows that 95% of companies are seeing zero return on their investments in AI technologies. Why? It turns out that AI's biggest bottleneck is not a technical one; it's a human one. As a result, U.S. businesses are missing out on an estimated $2.9 trillion in productivity gains not due to faulty or insufficient tools, but a lack of proper strategy and enablement. - Only about 20% of AI-driven projects meet their business objectives. - Firms are investing heavily in AI tools but 97 per cent of organisations still report a skills gap in their workforce. You cannot unlock productivity gains by deploying technology alone. Think about your own company. You bought ChatGPT Enterprise, Copilot, maybe Harvey or Cursor. How many employees actually use it daily to 10x their output? 5%? 10%? Most are extracting a fraction of the value. We have the smartest models in history sitting idle while middle managers still write status reports by hand. ![](https://storage.ghost.io/c/64/81/648165b7-979e-43dd-9c43-02101159f467/content/images/2026/08/5.jpg) ### What Would Actually Move GDP? If we actually want to grow our way out of debt with AI, we need to stop subsidizing H100s and start subsidizing *humans using AI*. 1. **Tax credits for adoption, not just R&D:** Today you get an R&D credit for building a model. You should get a productivity credit for retraining your accountant to close books in 2 hours instead of 2 days with AI. 2. **AI literacy as infrastructure:** Just like we built highways, we need national AI upskilling. Not "learn to code" but "learn to delegate to agents." 97% skills gap won't close itself. 3. **Outcome-based procurement:** Government shouldn't buy AI tools, it should buy AI outcomes. "Pay this contractor 20% more if they can process Medicare claims 50% faster using AI" - that's how you force adoption. GDP growth assumptions matter most. That assumption only holds if AI goes from 5% of workers extracting 10% value to 80% of workers extracting 80% value. Right now markets are focused on making models 10% smarter. The trillion-dollar question is making businesses 100% more adopted. Until we solve that, AI won't save the debt. It will just make Nvidia richer while interest payments keep outpacing growth. We don't need better models. We need more humans brave enough to actually use them.