AI Productivity and the Promised Land

Posted by Cam Hui - September 13, 2025

This is the first in a series of the opportunities and threats to productivity. I begin the series with a focus on the productivity effects of AI adoption.

There has been a lot of excitement over the productivity-enhancing promise of artificial intelligence. From a policy perspective, productivity is important as it defines the rate of non-inflationary growth potential of an economy.

The accompanying chart shows the evolution of U.S. total factor productivity during the post-war era, shown on a log scale. I focus on two periods, the PC revolution and the internet revolution. The IBM PC was introduced to the market in the early 1980s, though the technology was known to hobbyists in the 1970s. It wasn’t until the mid to late 1980s that they became ubiquitous and the economy began to see productivity gains from the PC revolution. Two black lines on the chart estimate the range of effects of PC revolution, which only produced relatively minor gains. The lower line is not much different from past trend growth, though the higher and steeper line shows the upper band of the productivity gains from this innovation. The internet revolution was different, as depicted by the solid red line, which shows a much steeper rate of growth that began in the early 2000s. I drew a dotted red line with the same slope as the internet revolution and superimposed it over the PC revolution period to show how much more the internet affected productivity during that period.

Looking forward, no one can accurately forecast the productivity gains from AI adoption. While different analysts can produce point forecasts, the error terms on the forecast are enormous. Will it be like the PC revolution, the internet or even better?

The AI Revolution

AI adoption is expected to revolutionize workflows in the coming years and the productivity gains that accompany the AI revolution. However, investors need to distinguish between how investors perceive AI and how AI affects the economy, and how the benefits of AI adoption are divided between all of the participants in the value chain.

Bob Elliott of Unlimited Funds estimated the GDP growth impact of AI for the next few years, and the effects are relatively modest. In reality, estimates of productivity growth from new technologies like AI are little more than guesses. The error term around estimates is very high relative to the level of the estimate itself.

At this point, AI-related productivity gains are nothing more than just educated guesses. What is more certain is the trend in job displacement from the AI revolution. A St. Louis Fed study found that job displacement is correlated with the degree of AI adoption across occupations. The study concluded:

Our results suggest we may be witnessing the early stages of AI-driven job displacement. Unlike previous technological revolutions that primarily affected manufacturing or routine clerical work, generative AI can target cognitive tasks performed by knowledge workers—traditionally among the most secure employment categories.

An article in the Economist outlined laid out a possible scenario of economic disruption:

Everyone else would have to adapt to gaps in AI’s abilities and to the spending of the new rich. Wherever there was a bottleneck in automation and labour supply, wages could rise rapidly. Such effects, known as “cost disease”, could be so strong as to limit the explosion of measured GDP, even as the economy changed utterly.

The new patterns of abundance and shortage would be reflected in prices. Anything AI could help produce—goods from fully automated factories, say, or digital entertainment—would see its value collapse. If you fear losing your job to AI, you can at least look forward to lots of such things. Wherever humans were still needed, cost disease might bite. Knowledge workers who switched to manual work might find they could afford less child care or fewer restaurant meals than today. And humans might end up competing with AIs for land and energy.

This economic disruption would be reflected in financial markets. There could be wild swings between stocks as it became clear which companies were winning and losing winner-takes-all contests. There would be a rapacious desire to invest, both to generate more AI power and in order for the stock of infrastructure and factories to keep pace with economic growth. At the same time, the desire to save for the future could collapse, as people—and especially the rich, who do the most saving—anticipated vastly higher incomes.

In summary, AI adoption promises high levels of benefits to productivity, but the economy will undergo a period of adjustment, and the error terms of productivity forecasts will be significant. One key question is the degree of job displacement from AI adoption. On one hand, hundreds of years of economic history shows that economies can remain at or near full employment as new technologies are introduced and adopted. On the other hand, the recent history of AI adoption indicates that most of the job displacement is in non-unionized non-blue-collar sectors, which is an uncharted waters dynamic that hasn’t been seen in the post-war history. The St. Louis Fed study summarized the problem facing forecasters:

The technology is so new that longer-term effects remain unknown. If generative AI drives sustained productivity growth, it could ultimately create new jobs and industries, potentially offsetting displacement effects. As AI capabilities continue advancing rapidly, monitoring these employment patterns becomes increasingly critical.

A Moat Breach?

I believe investor excitement over AI may be nearing a crescendo. Simon White at Bloomberg pointed

out that the aggregate capital expenditures of the largest AI firms is 1.3x its EBITDA, compared to a 50% rate of the companies in the rest of the S&P 100.

The moat of the leading AI plays is becoming eroded. The hyperscalers are all abandoning their capital light/intellectual property heavy business model as a source of competitive advantage. The pace of capital investment is turning the business models of the megacap technology companies upside down. Today, hyperscalers are engaged in capital spending that’s higher than old economy heavy capital-intensive companies. In other words, why are the likes of Alphabet and Amazon behaving like conventional industrial companies like GE and GM, at least in their capex policies?

Heavy capital spending would make sense in a winner-take-all world where the company with the best AI bot gains a dominant market position. But the evolution of Large Language Models (LLMs) shows that the law of diminishing returns is starting to become evident in their development. What’s the point of spending heavily on capex if all LLMs look similar to each other? Will hyperscalers end up like aircraft engine manufacturers, whose product lines are all finely engineered but appear to be commoditized products to the customers? Under that scenario, most of the added value of artificial intelligence would accrue to the end user.

The latest earnings report of semiconductor maker NVIDIA was revealing inasmuch as what management didn’t say. The company revealed that it had no H20 chip sales to China. This is an early sign that NIVIDA is struggling to maintain footing in China in the face of a surge by one of China’s own champions, Cambricon Technologies. I interpret this as an early sign that NVIDIA is starting to lose its competitive edge.

Sometimes in financial bubbles investors correctly identify the economic value of a technological innovation but misidentify who will capture that value. Market psychology conditions are signaling that frothy conditions are starting to come back down to earth. Consider this NY Times account of how Builder.ai went from a $1.5-billion valuation to zero within a few months. Remember Bob Farrell’s Rule #4: “Exponential rapidly rising or falling markets usually go further than you think, but they do not correct by going sideways.”

If the eclipse of hyperscaler dominance is starting, and the benefits of AI adoption depend on a new set of innovators who can exploit foundational models, Carl Benedikt Frey, writing in the IMF’s Finance & Development magazine, offered a bleaker outlook for productivity growth because industry concentration stifles innovation and growth:

Economic miracles stall when the institutions that enabled past successes become misaligned with new challenges. The Soviet Union and much of Europe stumbled when rigid mass production models failed to adapt to the unpredictability of the computer age, while Japan faltered as the epicenter of innovation shifted from hardware to software. Today, China’s growth is increasingly constrained by tightened party control, and the U.S. faces a similar peril whenever monopoly power remains unchecked. The danger that centralization and concentration will snuff out innovation now hangs over AI…

Beneath today’s veneer of intense competition, Microsoft’s deep alliance with OpenAI already controls about 70 percent of the commercial LLM market, while Nvidia provides about 92 percent of the specialized graphics-processing units (GPUs) used to train these models. Together with Alphabet, Amazon, and Meta, these incumbents have also been quietly buying stakes in promising AI start-ups. Sustaining a policy regime that safeguards the competitive arena itself, rather than the fortunes of particular firms, is essential if the next generation of transformative innovators is to deliver the promised boost to productivity. That’s as true for the AI age as it was for the computer era.

A Best-Case Scenario

To summarize the AI landscape today, David Sacks, tech founder and investor, and now White House special adviser for AI and crypto, recently outlined a “best-case scenario for AI” in a Twitter/X post:

The Doomer narratives were wrong. Predicated on a “rapid take-off” to AGI, they predicted that the leading AI model would use its intelligence to self-improve, leaving others in the dust, and quickly achieving a godlike superintelligence. Instead, we are seeing the opposite:

the leading models are clustering around similar performance benchmarks;

model companies continue to leapfrog each other with their latest versions (which shouldn’t be possible if one achieves rapid take-off);

models are developing areas of competitive advantage, becoming increasingly specialized in personality, modes, coding and math as opposed to one model becoming all-knowing.

None of this is to gainsay the progress. We are seeing strong improvement in quality, usability and price/performance across the top model companies. This is the stuff of great engineering and should be celebrated. It’s just not the stuff of apocalyptic pronouncements. Oppenheimer has left the building.

The AI race is highly dynamic, so this could change. But right now the current situation is Goldilocks:

We have five major American companies vigorously competing on frontier models. This brings out the best in everyone and helps America win the AI race. As @BalajiS has written: “We have many models from many factions that have all converged on similar capabilities, rather than a huge lead between the best model and the rest. So we should expect a balance of power between various human/AI fusions rather than a single dominant AGI that will turn us all into paperclips/pillars of salt.”

So far, we have avoided a monopolistic outcome that vests all power and control in a single entity. In my view, the most likely dystopian outcome with AI is a marriage of corporate and state power similar to what we saw exposed in the Twitter Files, where “Trust & Safety” gets weaponized into government censorship and control. At least when you have multiple strong private sector players that gets harder. By contrast, winner-take-all dynamics are more likely to produce Orwellian outcomes.

There is likely to be a major role for open source. These models excel at providing 80–90% of the capability at 10–20% of the cost. This trade-off will be highly attractive to customers who value customization, control and cost over frontier capabilities. China has gone all-in on open source, so it would be good to see more American companies competing in this area, as OpenAI just did (Meta also deserves credit).

There is likely to be a division of labour between generalized foundation models and specific verticalized applications. Instead of a single superintelligence capturing all the value, we are likely to see numerous agentic applications solving “last mile” problems. This is great news for the start-up ecosystem.

There is also an increasingly clear division of labour between humans and AI. Despite all the wondrous progress, AI models are still at zero in terms of setting their own objective function. Models need context, they must be heavily prompted, the output must be verified and this process must be repeated iteratively to achieve meaningful business value. This is why Balaji has said that AI is not end-to-end but middle-to-middle. This means that apocalyptic predictions of job loss are as overhyped as AGI itself. Instead, the truism that “you’re not going to lose your job to AI but to someone who uses AI better than you” is holding up well.

In summary, the latest releases of AI models show that model capabilities are more decentralized than many predicted. While there is no guarantee that this continues — there is always the potential for the market to accrete to a small number of players once the investment super-cycle ends — the current state of vigorous competition is healthy. It propels innovation forward, helps America win the AI race and avoids centralized control. This is good news, which the Doomers did not expect.

Cam here. I agree with most of Sacks’ comments, but keep in mind that he is outlining a best-case scenario. Foundation model development is mostly finished. We believe it may be time to look for AI 2.0 investment plays, or industries that are likely to capture more of the value of the AI revolution. Some sectors, such as healthcare, will remain labour intensive and less sensitive to the productivity-enhancing benefits of AI adoption. Manufacturing and consumer-facing businesses could see significant productivity gains if AI is adopted properly. On the other hand, China is catching up quickly in AI, not only in open-source modeling but also in semiconductors, which will erode the U.S. competitive advantage in setting a global standard. In a related field, China appears to have leapfrogged other major economies in robotic development and applications.

Analysts will have to keep in mind the scenario of a lower productivity path outlined by Carl Benedikt Frey. Growing industry concentration stifles productivity and innovation. Incumbents become more interested in exploiting their dominant position to lobby for greater regulatory advantage such as patent extensions, tie up human capital with non-compete agreements and asserting power through licensing models.

In conclusion, there has been a lot of excitement over the productivity-enhancing promise of artificial intelligence. Investors need to distinguish between how investors perceive AI and how AI affects the economy, and how the benefits of AI adoption are divided between all of the participants in the value chain. While the AI revolution is very real, investor excitement over AI may be nearing a crescendo. It may be time to look for AI 2.0 investment plays, or industries that are likely to capture more of the value of the AI revolution.

2 thoughts on “AI Productivity and the Promised Land”

David Lynam says:

September 13, 2025 at 9:41 am

If (Microsoft.Nvdia) together with Alphabet, Amazon, and Meta, … have also been quietly buying stakes in promising AI start-ups…It may be time to look for AI 2.0 investment plays, or industries that are likely to capture more of the value of the AI revolution. Then why do investors need to look any further than the big 5?
1. Cam Hui says:
  
  September 13, 2025 at 10:10 am
  
  Recall that during the dot-com bubble Microsoft and other major tech companies used their overpriced stock as a medium of exchange to make acquisitions. How did those acquisitions turn out and how did the market turn out?

Comments are closed.