As of November 30, 2025, there is strong early evidence in the direction of Friedberg’s thesis, but the claim is explicitly about where future AI-driven productivity and value will ultimately accrue, and the time horizon (likely late‑2020s/2030s) has not yet arrived.

1. Are traditional industrial sectors seeing meaningful AI-driven gains yet?
Yes, especially in manufacturing and pharma, but these gains are still in the early scaling phase rather than a settled end state:

• Manufacturing: Recent industry surveys report very high AI adoption and large reported impacts—e.g., up to ~63% productivity boosts and ~20% cost reductions in plants that have implemented AI at scale, with widespread use in predictive maintenance, visual inspection, and smart supply chains. (zipdo.co) Other data show AI integrated into many factories’ robotics and control systems, with projections of ~20% productivity growth in manufacturing by 2035 and large cost savings from predictive maintenance. (gitnux.org) Most manufacturers say AI is already improving quality and decision‑making and expect it to materially boost revenue over the next few years. (manufacturingtomorrow.com)
• Medicine/pharma: Think‑tank and consulting reports find AI can plausibly cut drug‑development timelines roughly in half and add 2.6–4.5% of annual revenues in pharma and medical products, with pharma companies currently capturing about half of the generative‑AI‑in‑drug‑discovery market’s revenue. (itif.org) Large incumbents like Eli Lilly are now building AI supercomputing partnerships specifically to accelerate discovery and manufacturing, reflecting a serious bet that AI will materially improve productivity and time‑to‑market. (reuters.com)
  Overall, this supports the direction of his view that traditional industrial sectors can get large productivity gains from AI, software, robotics, and automation, but much of the quantified impact is still framed as 2030+ potential rather than already realized.

2. Is “much of AI’s economic value” clearly accruing to these sectors instead of pure‑tech companies yet?
Here the picture is mixed and clearly not settled:

• On the one hand, broad surveys show that only a small minority of firms globally (about 5%) are getting measurable value from AI so far, and the industries with the highest AI maturity and realized value today include software, telecom, and fintech, while many traditional sectors (e.g., chemicals, construction, real estate) lag. (businessinsider.com) This suggests that, at least in the 2023–2025 window, realized AI value is still concentrated in tech and a handful of more advanced adopters rather than being clearly dominated by traditional industrial incumbents.
• Equity‑market evidence shows that the largest and most obvious financial beneficiaries of the AI boom to date are mega‑cap technology and semiconductor companies like Nvidia, Microsoft, Alphabet, and others, whose market caps have surged to multi‑trillion‑dollar levels largely on AI infrastructure and platform economics. (euronews.com) At the same time, many AI‑native startups are struggling to convert hype into revenue, so value within tech is skewed to large incumbents rather than new entrants, but it is still overwhelmingly tech‑sector value, not clearly dominated by food, manufacturing, or pharma stocks. (investor.wedbush.com) Because Friedberg’s prediction concerns where the bulk of AI’s economic value will ultimately accrue—a structural, long‑run outcome—not just who is winning in the first 1–2 years of the boom, current data on market caps and early adoption do not yet suffice to judge him right or wrong.

3. Why this is “inconclusive (too early)” rather than right/wrong/ambiguous

• Too early relative to the horizon: The core of the claim is explicitly forward‑looking (“will derive substantial future gains” and “that’s where I’d concentrate capital”). Most rigorous analyses of AI in manufacturing and medicine frame the main productivity and economic impact in the late‑2020s and 2030s, not by 2025. (gitnux.org)
• Evidence is directionally supportive but not decisive: We already see meaningful pilot‑ and early‑scale productivity gains in industrial sectors, consistent with his thesis, but they do not yet amount to clear evidence that these sectors are capturing most of AI’s economic value compared with tech platforms.
• Outcome metrics are inherently long‑run: Determining whether "much" of AI’s eventual economic value accrues to traditional industrial incumbents versus pure‑tech firms will require observing several more years of diffusion, commoditization of core models, and integration into physical production and healthcare systems.

Because both the productivity gains and the cross‑sector value distribution he describes are still unfolding on a longer time frame than the ~20 months since the podcast, the prediction cannot yet be fairly scored as right or wrong; it is best classified as inconclusive (too early).