Edward Conard

Top Ten New York Times Bestselling Author

  • “…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
  • “Unintended Consequences is far smarter and more thought-provoking than most economics written for the general public” - Greg Mankiw, Harvard University, Former Chairman of the Council of Economic Advisors
  • “Unintended Consequences offers deep and well-argued analyses on almost every issue.” - The New York Times
  • “There are an amazing number of good ideas and interesting points made in Unintended Consequences. The thinking underlying it, and the obvious depth of understanding of the author, are very impressive.” - Steven Levitt, coauthor of Freakonomics; 2004 John Bates Clark Medal
  • “…a must-read for serious students of economic policy.” - Glenn Hubbard, Dean, Columbia Business School, and former Chairman of the Council of Economic Advisers
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “Unintended Consequences should be read by anyone who takes for granted the superiority of progressive taxation and has not thought carefully about the trade-offs involved.” - The New Republic
  • “Unintended Consequences provides a provocative interpretation of the causes of the global financial crisis and the policies needed to return to rapid growth. Whether you agree or not, this analysis is well worth reading.” - Nouriel Roubini, New York University; Chairman, Roubini Global Economics
  • “…a very valuable contribution.” - Larry Summers, former Secretary of the Treasury and director of the National Economic Council, president emeritus, Harvard University
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “A full-throated defense of economic dynamism.” - The Wall Street Journal
  • “…reminds us that inequality sends a signal of what society lacks most, in America’s case, entrepreneurship and risk taking.” - Lawrence Lindsey, CEO, The Lindsey Group, former Director of the National Economic Council
Upside of Inequality Oxford Unintended Consequences
Buy the Books
  • Macro Roundup
  • About Roundup
  • About Ed Conard
  • Highlights
  • Topics
  • Subscribe
Edward Conard
  • twitter
  • facebook
  • linkedin
  • youtube
  • Email
  • Text Message (SMS)
  • Twitter/X
  • LinkedIn
  • Facebook
  • WhatsApp Message
Subscribe to Macro Roundup Emails
  • Mentions 374
  • Primary focus 78
For whatever topics you select (currently: Innovation/Research):
Choose search scope

Your importance filter 'Database' shows fewer articles.

Remove filters to see full article counts

The Dream of Bringing Back Bell Labs

AI Summary. Google's artificial intelligence research units have been an important driver of machine learning research, and large private firms including IBM are active in quantum computing research. Some large startups such as SpaceX conduct in-house research that pushes the boundaries of general-purpose technology rather than pursuing quick profits.

Noah Smith Noahpinion
Date Posted:
December 8, 2022
Is Database:
Database

.@Noahpinion reviews the shift of American innovation from corporate labs to university research and startups but notes that Google’s AI divisions have been an important driver of research in the frontier of machine learning innovation.

Are large private companies still driving frontier technology research?

Core argument: The mid-20th-century successes of Bell Labs and other large corporate research labs have led many observers to argue that such labs are an important missing piece of the modern innovation ecosystem.

The successes of Bell Labs and other big corporate labs in the mid 20th century has many people thinking that maybe this is an important missing piece of our modern innovation ecosystem. Google’s AI divisions have been an important driver of research in the machine learning space — an extremely important frontier. All told, the research output of Google AI, Google Brain, DeepMind, etc. has been truly staggering: Big private companies (especially IBM) are also very active in quantum computing research. And some “startups” like SpaceX are big enough to do research in-house that pushes the boundaries of general-purpose technology instead of just making a quick buck.

Takeaways by Macro Roundup® AI

  1. The mid-20th-century successes of Bell Labs and other large corporate research labs have led many observers to argue that such labs are an important missing piece of the modern innovation ecosystem.
  2. Google's AI units, including Google AI, Google Brain, and DeepMind, have been an important driver of machine learning research, with a research output described as staggering.
  3. Large private companies, especially IBM, are active in quantum computing research, and companies such as SpaceX are big enough to conduct in-house research that pushes the boundaries of general-purpose technology.

Noah Smith reviews the shift of American innovation from corporate labs to university research and startups but notes that Google’s AI divisions have been an important driver of research in the extremely important frontier of machine learning innovation, “The successes of Bell Labs and other big corporate labs in the mid 20th century has many people thinking that maybe this is an important missing piece of our modern innovation ecosystem. Google’s AI divisions have been an important driver of research in the machine learning space — an extremely important frontier. All told, the research output of Google AI, Google Brain, DeepMind, etc. has been truly staggering: Big private companies (especially IBM) are also very active in quantum computing research. And some “startups” like SpaceX are big enough to do research in-house that pushes the boundaries of general-purpose technology instead of just making a quick buck.“

Noah Smith, “The dream of bringing back Bell Labs,” Noahpinion, December 8, 2022, https://noahpinion.substack.com/p/the-dream-of-bringing-back-bell-labs

The dream of bringing back Bell Labs

I just finished reading The Idea Factory: Bell Labs and the Great Age of American Innovation, by Jon Gertner. It’s a fun and fascinating book chock full of stories about the invention of the vacuum tube, the transistor, information theory, communications satellites, fiber-optic cable, and so on. The personalities really come to life — the pragmatic Mervin Kelly, the arrogant and paranoid William Shockley, the obsessive and curious Claude Shannon, the distractible and creative John Pierce. I highly recommend it.

That said, the book is also a bit frustrating, because it grasps at a general theory of collective innovation, but never quite reaches a conclusion. How much of breakthrough discovery is due to the team vs. the individual? How do incentives and organizational structure matter? How much of the success of Bell Labs was due to the luck of history versus the unique way the organization was set up? The scientists at Bell themselves grappled with these questions, but nobody has ever answered them conclusively.

The way America innovates has certainly changed since the mid 20th century. In a 2019 paper, the economists Ashish Arora, Sharon Belenzon, Andrea Patacconi and Jungkyu Suh document two major trends: the rise of university research, and the decline of corporate labs. Basically, businesses still spend a ton on R&D, but more and more of what they spend goes to development (getting innovations ready for market) than research itself.

Bell Labs is one of the iconic corporate labs, along with Xerox PARC, DuPont’s labs, a few others, whose decline illustrates the general trend.

The authors document two main reasons for the change. First, the government started supporting universities a lot, both with money, and with legal changes (especially the Bayh-Dole Act) that allowed universities to profit from licensing their research to private companies. Second, the maturation of the startup and venture capital ecosystem, increased competition, and a more tolerant government attitude towards acquisitions all pushed big companies to buy innovations rather than producing them in-house. Note the ambiguous effect of antitrust here — stronger antitrust promotes competition (as when Bell was broken up in 1982), while weaker antitrust lets companies substitute M&A for research.

The net result of these changes was to create a sort of supply chain for innovation — universities do the scientific research, startups and university spinoffs and mission-oriented government agencies like DARPA turn insights into inventions, and big companies acquire the inventions and focus on turning them into products. Here’s how the authors sum it up: “In summary, the new innovation ecosystem exhibits a deepening division of labor between universities that specialize in basic research, small start-ups converting promising new findings into inventions, and larger, more established firms specializing in product development and commercialization…Large firms therefore invest in scientific capability not so much to generate knowledge as to be effective buyers of knowledge.”

From the standpoint of economic theory, this sort of supply chain makes a lot of sense. Research discoveries and fundamental inventions are public goods, because they have a strong tendency to leak out across organizational boundaries, making it impossible for whoever did the research to capture the monetary value. The transistor was invented (discovered?) at Bell Labs, by John Bardeen, Walter Brattain, and William Shockley; how much of the total monetary value created by transistors was captured by those individuals, their research teams, or Bell itself? Very, very little of it. This means that private companies have very little monetary incentive to do basic research, which is why economists generally think basic research should be left to universities and the government (or, depending on which economists you ask, to quixotic rich people).

Arora et al. sound a note of caution, though. They point out several attractive features of big corporate labs that the new innovation supply chain might lack. First of all, while many university researchers focus on questions of general scientific interest (how the Universe was formed, for instance), corporate labs tended to focus on the creation of general-purpose technologies — things that had lots of pragmatic industrial uses, like transistors. The inflationary theory of cosmology and the invention of the transistor are both “basic research”, but their economic value is very very different. Second, corporate labs tend to be multi-disciplinary, pulling in researchers from a wide variety of fields and having them work together on projects, while university research — as anyone who has worked in academia knows all too well — tends to be very siloed by discipline.

For these reasons, universities may not be producing the research that would have the biggest economic benefit. Now, you might not care about this — after all, learning where the Universe came from is pretty cool! But at some point we have to face the fact that we’re spending more and more on research but getting only about the same amount of total economic benefit for all that spending:

At some point we should probably think about increasing the economic productivity of scientific research. The successes of Bell Labs and other big corporate labs in the mid 20th century has many people thinking that maybe this is an important missing piece of our modern innovation ecosystem.

Not completely missing, though. As Arora et al. point out, Google’s AI divisions have been an important driver of research in the machine learning space — an extremely important frontier. All told, the research output of Google AI, Google Brain, DeepMind, etc. has been truly staggering:

Big private companies (especially IBM) are also very active in quantum computing research. And some “startups” like SpaceX are big enough to do research in-house that pushes the boundaries of general-purpose technology instead of just making a quick buck. So Arora et al. are probably overstating the changes to America’s innovation structure a bit; as their own data shows, businesses still do quite a bit of research, even if not as much as before.

But could we bring back something as amazing as Bell Labs? Many people argue that Bell’s success was due to very unique conditions that we probably won’t see ever again. I’ll outsource this argument to Ilan Gur, the CEO of the UK’s science innovation agency ARIA, who laid it out in an interesting Twitter thread last January:

Gur argues that there’s just too much well-funded and widely distributed competition for research talent and discoveries these days for an organization like Bell Labs to dominate the landscape. He suggests that instead, we should focus on helping startups do deep fundamental innovation, like Moderna and BioNTech did with mRNA vaccines.

It’s a reasonable thesis, and I think there’s lots of value in deep-tech “super-startups” (a term I just invented) that put tons of money into trying to do fundamental research into mRNA, reusable rockets, aneutronic fusion, generative AI, solid-state Li-ion batteries, Crispr, and any number of other fundamental technologies. But there’s a natural limitation here, which is the whim of the market itself. We just saw a massive decline in tech stock prices. This led to a big crunch in startup funding, especially for expensive late-state startups. So far, software companies seem to have taken the brunt of the current crash, while deep-tech companies are doing OK. But given the natural difficulty of capitalizing on fundamental research discoveries (there’s that pesky public goods problem again!), it seems likely that deep-tech super-startups whose efforts will take many years to see fruition will be in the firing line at some point.

So I think there might still be a role for more Bell-style corporate labs in fields where it seems like there’s lots of important breakthrough research left to be done. In The Idea Factory, Gertner suggests two possible such fields: biotech, and energy. Let’s consider the second of these two.

The energy transition is one of the most important events of our time, both because of the need to avert climate catastrophe and because of the promise of ultra-cheap energy. Right now, while startups are trying to do some basic research in energy tech, most of the effort is being done by universities, government labs, or mission-focused government agencies like ARPA-E. But what if there was a Bell Labs for energy tech?

Google can fund AI research with the money from its quasi-monopoly in search advertising. But who would fund a giant high-class world-beating private lab system to do fundamental energy research? There are no big monopoly companies in the energy space — our big energy companies are all fossil fuel companies, and they’re spending their efforts on slowing the energy transition instead of speeding it up.

So here’s a wacky idea: What about a national electrical utility? Many utilities are already government-sanctioned monopolies at the city level, so the principle of natural monopolies in the energy space is well-established. And the intermittency of wind and solar means it might make sense to run new long-distance power lines between cities. So why not turn this task over to a national-level government-sanctioned utility monopoly, that would own the long-distance transmission lines and would also be allowed to invest in local generation capacity?

In fact, Bernie Sanders already suggested creating a government agency to do this, but that’s probably a political non-starter and would have questionable incentives as well. A sanctioned national utility company — call it Energy Bell — with profit margins limited by law, would be motivated by the profit motive — but with capped margins, the only way it could make profits would be to deliver American customers greater volumes of energy. So just as Bell’s driving mission was to connect the whole world, a monopoly energy company’s mission would be to provide more abundant energy to more Americans (with laws to make sure it’s mostly or entirely clean energy).

Meanwhile, the comfy position of a sanctioned monopoly would give Energy Bell the long time horizon to establish its own big corporate research lab, focused on solving deep fundamental scientific problems related to energy — solid-state batteries, batteries that don’t use lithium, aneutronic fusion, seasonal energy storage, solar panel and battery recycling, or whatever. And just to provide a little extra incentive, doing this research could be a condition of Energy Bell’s continued existence, written right into its charter.

And Energy Bell’s large scale could give it the money to hire many of the world’s best researchers in a number of disciplines and put them to work in multidisciplinary, cooperative ways. If we allowed Energy Bell to ignore the cap on employer-sponsored green cards, it would be even easier to assemble this super-team.

Anyway, I realize this is a wacky idea, and very unlikely to happen. We’re in an era of rising distrust of big monopolies, and lots of people in both industry and politics would likely resent a company like Energy Bell. On top of that, there are probably a bunch of technical problems with the idea that I haven’t yet thought of. But if we wanted to bring back something akin to the old Bell Labs, in an incredibly crucial area of fast-moving technology, I think this would be a promising way to do it.

Barring something like an Energy Bell, though, I agree with Ilan Gur that we’re unlikely to see something quite like Bell Labs in our lifetimes. The university-DARPA-startup innovation system is probably here to stay, and we should focus on finding ways to make this technological supply chain work in a more efficient, purposeful, and well-integrated way.

  • Innovation/Research
  • Productivity
    • Institutional Capabilities
Previous articleDecember 8, 2022Unions’ Inflation Warning?.@EthanYWu at @ft shares Blackrock analysis forecasting declining American labor force participation rates driven by aging. This implies that economic activity “will need to run at a lower level to avoid persistent wage and price inflation,”Next articleDecember 8, 2022Beijing Running Out of Fever Medication as Covid Outbreak SpreadsIbuprofen is being rationed in Beijing, and new modelling suggests that as many as 1M people could die in China in a “winter wave” in the coming months. @ft
Showing 78 database articles primarily about Innovation/Research
Currently filtering by:
  • Remove 'Database'

No Signs of AI Payoff Outside the Tech Sector

AI Summary. Revenue per worker and profit margins are rising for the seven largest technology companies but flat for the rest of the S&P 500, with no signs of artificial intelligence boosting productivity growth. The author argues that as long as spending on artificial intelligence stays confined to a small part of the economy, its macroeconomic effects will remain very limited.

Torsten Slok Apollo
Date Posted:
October 26, 2026
Is Database:
Database

Citing flat revenue per worker, and profit margin growth for the S&P 493, Slok argues that so far, AI’s diffusion through the economy has delivered “zero financial payoff to corporate America outside of tech.”

Is artificial intelligence boosting revenue and margins outside big tech?

Core argument: Revenue per worker and profit margins are rising for the Magnificent 7 but flat for the S&P 493, with no signs of AI boosting productivity growth or delivering financial payoff outside the tech sector.

With no impact on the bottom line, mainstream corporate America is pulling back and starting to view token spending not as a source of growth but as a cost center that weighs on margins. VC-backed companies are spending more and more on tokens, but mainstream corporate America does not need the latest model and is unwilling to pay for it. As long as agentic AI and token spending are confined to a small part of the economy, the macroeconomic effects of AI will be very limited. The question looking ahead is whether, even as models become commoditized, mainstream corporate America can turn increasingly cheap intelligence into higher revenue growth and higher profit margins. That may be harder than it sounds. The AI story will be decided in the S&P 493, not the Mag 7, and so far, there is no evidence that revenue growth or margins are improving outside the tech sector.

Takeaways by Macro Roundup® AI

  1. Revenue per worker and profit margins are rising for the Magnificent 7 but flat for the S&P 493, with no signs of AI boosting productivity growth or delivering financial payoff outside the tech sector.
  2. Mainstream corporate America is pulling back from AI and treating token spending as a cost center that weighs on margins, while VC-backed companies spend increasingly more on tokens.
  3. The macroeconomic effects of AI will remain very limited as long as agentic AI and token spending stay confined to a small part of the economy.

Related Articles:

  • Circularity in AI Shows Up in Token Spending Too — Payment-data analysis finds 80% of the largest corporate AI spenders are venture-backed, versus 13% of the median spender, indicating that much token spending is funded by venture capital rather than earnings. If AI payoffs are slow, venture returns could suffer and a funding pullback could weaken demand for computing power.
  • Califoreboding — A basket of 42 AI stocks accounts for 60%-80% of all S&P 500 earnings growth, price returns and capital spending, while 60% of S&P 500 stocks sit 20% or more below their highs and only 30% outperform the index. AI-related construction spending has risen 38%, the only positive trend as manufacturing, commercial real estate and residential construction decline.
  • A Catalyst For The AI Bubble Break — Historically, controlling for valuations, a 1% rise in the share of the market represented by initial public offerings has been associated with roughly a 4% decline in prices. With large artificial-intelligence company listings expected and share issuance exceeding buybacks, equity supply could shift from shrinking about 1% a year to growing near 5%, implying returns roughly 20% below average.
  • Innovation/Research
  • Productivity

The College Wage Premium in the Generative AI Era

AI Summary. The U.S. college wage premium fell from 0.626 to 0.575 after four decades of expansion, implying the first sustained negative growth in relative demand for college-educated labor. The authors estimate that weaker wage growth in occupations highly exposed to artificial intelligence accounts for roughly 28% of that decline, though the patterns are not causal.

José Azar, Mireia Gine and Javier Sanz-Espín Social Science Research Network
Date Posted:
September 4, 2026
Is Database:
Database

The college wage premium flattened in the mid-2010s and has fallen ~8% since 2022. The authors argue that this compression reflects a broad decline in the returns to formal schooling, rather than a decline in the upper tail.

Is AI exposure driving the decline in the college wage premium?

Core argument: The U.S. college wage premium fell from 0.626 in 2022 to 0.575 in 2026 after expanding for four decades, with Current Population Survey data implying the first sustained negative relative demand growth for college labor.

After expanding for four decades, the U.S. college wage premium [dropped] sharply from 0.626 in 2022 to 0.575 in 2026. Current Population Survey data through 2026 implies an unprecedented drop in relative demand for college labor—the first sustained negative relative demand growth. Post-2022 wage growth slowed disproportionately in high-exposure occupations, which employ a disproportionate share of college graduates. By 2026, going from zero occupational AI exposure to full exposure had a negative effect on wages of−0.086. Combined with the college–non-college exposure gap, this mechanism accounts for roughly 28% of the total drop in the college wage premium from 2022 to 2026. While non-causal, these patterns indicate that task displacement in AI-exposed white-collar occupations plays a quantitatively meaningful role in the recent compression of the aggregate skill premium.

Takeaways by Macro Roundup® AI

  1. The U.S. college wage premium fell from 0.626 in 2022 to 0.575 in 2026 after expanding for four decades, with Current Population Survey data implying the first sustained negative relative demand growth for college labor.
  2. By 2026, moving from zero occupational AI exposure to full exposure had a -0.086 effect on wages, and combined with the college–non-college exposure gap this mechanism accounts for roughly 28% of the 2022–2026 drop in the college wage premium.
  3. Post-2022 wage growth slowed disproportionately in high-exposure occupations that employ a disproportionate share of college graduates, patterns that, while non-causal, indicate task displacement in AI-exposed white-collar occupations plays a quantitatively meaningful role in compressing the skill premium.

Related Articles:

  • Looking for the Ladder — The article argues that the hiring downturn in occupations most exposed to generative AI reflects interest-rate sensitivity rather than AI itself, since roughly 38% of workers in the most AI-exposed quintile work in information, finance and insurance, and professional and technical services, versus under 2% in the least-exposed quintile.
  • How Students and Recent Grads are Responding to the Rise of AI — Undergraduate enrollment in degrees tied to the most AI-exposed jobs has risen 8%, indicating students are moving toward rather than away from fields exposed to artificial intelligence. A decline in computer science degrees is more than offset by growth in other exposed fields such as engineering.
  • AI and Young-adult Jobs: The Real Mystery — Employment rates have declined for young adults both with and without college degrees, with non-degree holders faring marginally worse, and employment growth for young workers of all education levels has lagged that of other workers. Artificial intelligence can replace some lower-skilled jobs, but those jobs are too few to drive youth labor market weakness.
  • Innovation/Research
  • Productivity
  • Workforce
    • Education
      • College
    • Unemployment/Participation

The AI Re-Acceleration That Wasn’t

AI Summary. An analysis of Epoch's Capabilities Index data, controlling for developer and model family, finds no statistically significant breakpoint in AI progress; a segmented model fits no better than a straight-line trend (p = 0.615), with the estimated slope change ranging from -8.4 to +23.4 points per year.

Paul Kedrosky Applied Complexity
Date Posted:
September 3, 2026
Is Database:
Database

Kedrosky argues AI capabilities continue to improve, but “the full composite data shows flattening relative gains, not acceleration…rolling relative model gains have fallen from their 2024 peak, while model dispersion has narrowed sharply.”

Has AI capability progress accelerated beyond its linear trend?

Core argument: Using all observations from Epoch's Capabilities Index and controlling for developer and model family, a piecewise model shows no statistically significant breakpoint in AI capability trends (p = 0.615) and does not improve on a purely linear trend.

Using all Epoch’s Capabilities Index observations, and controlling for developer and model family, there is no statistically significant breakpoint. A piecewise model—which splits the series into intervals and applies a sub-function to each segment—does not improve on a purely linear trend: p = 0.615, The estimated change in slope has a confidence interval of -8.4 to +23.4 points per year. In short, the maths shows there is no model acceleration, contrary to claims, and as expected. The result comes from selecting frontier observations only, choosing a breakpoint, ignoring variance collapse, and fitting separate lines on either side.

Takeaways by Macro Roundup® AI

  1. Using all observations from Epoch's Capabilities Index and controlling for developer and model family, a piecewise model shows no statistically significant breakpoint in AI capability trends (p = 0.615) and does not improve on a purely linear trend.
  2. The estimated change in slope of the AI capabilities trend carries a confidence interval of -8.4 to +23.4 points per year.
  3. Apparent AI acceleration results from selecting frontier observations only, choosing a breakpoint, ignoring variance collapse, and fitting separate lines on either side of that breakpoint.

Related Articles:

  • Why .400 Hitters Disappeared — and What It Means for AI — Artificial intelligence models' capability gains are shrinking in percentage terms, falling from about 15% relative to the baseline early on to roughly 10% for the same absolute increment, narrowing the gap between frontier and good-enough models. The author argues this points to a commoditizing market where price, not minor model differences, becomes the main differentiator, pressuring frontier firms' margins.
  • Chart of the Day: Small Models are Closing the Gap to Frontier AI — Achieving 60% on MMLU, a test of academic reasoning, once required a 540bn-parameter model but was reached by a 3.8bn-parameter model two years later, a 142x reduction in parameters. Models small enough to run on phones or sensors shift inference from data centers to devices, weakening the case for paying cloud providers per token.
  • Anthropic’s Best AI Model Struggles To Attract Users As Cheaper Tools Thrive — Spending on Anthropic's largest and priciest AI model has plateaued at about 11% of total outlay on the company's tools, according to payments data from 70,000 companies, breaking a pattern of corporate users defaulting to the most powerful models. Analysts and investors cite its high price and older models' adequacy.
  • Innovation/Research
  • Productivity
    • Investment

The AI Trade Is Losing One of Its Key Signals

AI Summary. An index blending artificial intelligence token prices and usage is ambiguous when it softens, since token prices have fallen more than 90% since 2023 while total spending has roughly doubled over the past year. Allianz Research measures a nearly 46% growth gap between artificial intelligence investment and sales, wider than the 32% divergence during the 2001 telecom bust.

Jan-Patrick Barnert and Michael Msika Bloomberg
Date Posted:
July 7, 2026
Is Database:
Database

The Silicon Data LLM Token Expenditure Index, which tracks AI token prices, is down ~20% from its recent peak in May. The metric is a rough “proxy for marginal willingness to pay.”

Why does a softening AI token index send ambiguous signals?

Core argument: Token prices have fallen more than 90% since 2023 while total token spend has roughly doubled over the past year, supporting a benign read in which cheaper tokens expand the market and an index pause reflects digestion.

A softer index doesn’t mean AI is getting cheaper. The gauge blends prices and usage, meaning a dip can imply very different scenarios: either list prices are falling, or demand is shifting toward cheaper models. It could also point to a genuine softening in what buyers are prepared to shoulder. Each of these possibilities carries different implications. Let’s start with a benign read: While token prices have collapsed more than 90% since 2023, total spend has roughly doubled since last year. Cheaper tokens have expanded the market. This means that an index pause is simply digestion, while demand is real and capex is money well spent. The bull case for Nvidia rests here. Now for the interpretation that’s keeping people up at night: Bears warn that sustained weakness in the index could end the trade that saw nearly the entire AI cohort rally hard this cycle. It’s token spending that justifies the next capex order, and the bill is already looking stretched. Allianz Research said there’s nearly a 46% growth gap between AI investment and sales. That’s worse than the 32% divergence measured during the 2001 telecom bust.

Takeaways by Macro Roundup® AI

  1. Token prices have fallen more than 90% since 2023 while total token spend has roughly doubled over the past year, supporting a benign read in which cheaper tokens expand the market and an index pause reflects digestion.
  2. Allianz Research put the growth gap between AI investment and sales at nearly 46%, wider than the 32% divergence measured during the 2001 telecom bust.
  3. The index blends prices and usage, so a dip can imply falling list prices, demand shifting toward cheaper models, or a genuine softening in what buyers are prepared to shoulder.

Related Articles:

  • AI: The ROI Runway Could Be Long Outside the Tech Sector — AI company valuations rest on the assumption that profit margins across non-tech firms will eventually rise, but outside a handful of sectors adoption is slow, creating a divergence between front-loaded valuations and slower cash flows. If the per-use costs of AI models converge toward zero, revenue may be insufficient for all major cloud providers.
  • Semiquincententacles — Leading artificial-intelligence labs are moving to usage-based token pricing on a cost-plus basis that includes total ownership costs, pushing token prices higher and giving some enterprises reason to migrate to cheaper open models. Frontier models remain necessary for cybersecurity, distillation training, agents, and scientific discovery, where maximum performance is required.
  • Chart of the Day: Small Models are Closing the Gap to Frontier AI — Achieving 60% on MMLU, a test of academic reasoning, once required a 540bn-parameter model but was reached by a 3.8bn-parameter model two years later, a 142x reduction in parameters. Models small enough to run on phones or sensors shift inference from data centers to devices, weakening the case for paying cloud providers per token.
  • Innovation/Research
  • GDP
    • Financial Markets
  • Productivity
    • Investment

The Age Of The Solopreneur

AI Summary. A surge in US business formation is being driven by solo entrepreneurs, with growth in AI tool usage and in solopreneur activity in high-AI-adoption sectors suggesting artificial intelligence accounts for a meaningful portion of the increase. The authors argue AI is lowering barriers to starting and growing a business by expanding what one-person firms can do.

Ernie Tedeschi, Marisa Rama and Chris Cruickshank Stripe Economics
Date Posted:
June 22, 2026
Is Database:
Database

New US business applications have surged since late 2024, driven mainly by “likely nonemployer” filings representing “solopreneurs” rather than businesses expected to hire. Data suggest AI may be enabling this by filling capability gaps without hiring.

Is AI driving the surge in one-person business formation?

Core argument: The recent surge in US business formation is not reflected in high-propensity applications, but the evidence points toward a structural increase in genuine small business activity over 2025 and 2026, driven by solopreneurs.

While the recent surge in US business formation is not reflected in high-propensity applications, the evidence points toward a structural increase in genuine small business activity over 2025 and 2026, driven by solopreneurs. Preliminary evidence of growth in AI tool usage and solopreneur growth in high-AI-adoption sectors suggests that advances in AI are responsible for a meaningful portion of this growth. We believe AI is lowering barriers to business formation and growth by expanding the capabilities of solopreneurs, further improving tools and platforms that cater to new businesses, and creating a new set of opportunities for entrepreneurs to pursue. Whether the magnitude of this effect is as large as the most optimistic readings of the data suggest remains to be seen, but we believe we might be in the early innings of a fundamental acceleration in business formation—which could have ripple effects throughout the economy.

Takeaways by Macro Roundup® AI

  1. The recent surge in US business formation is not reflected in high-propensity applications, but the evidence points toward a structural increase in genuine small business activity over 2025 and 2026, driven by solopreneurs.
  2. Preliminary evidence of growth in AI tool usage and solopreneur growth in high-AI-adoption sectors suggests that advances in AI are responsible for a meaningful portion of this growth.
  3. AI appears to be lowering barriers to business formation and growth by expanding solopreneur capabilities, improving tools and platforms serving new businesses, and creating opportunities for entrepreneurs, with possible ripple effects throughout the economy.

Related Articles:

  • Solopreneurs, Solow, and the SaaSpocalypse — For three decades after commercial electricity delivery began in 1882, inflation-adjusted output per worker grew just 0.5% annually, accelerating to more than double that pace only after factories were redesigned around electric motors rather than retrofitted from steam. The author argues a near-term absence of productivity acceleration from AI is not evidence none is coming.
  • AI-Native Firms∗ — The authors find AI-native U.S. startups are 25% smaller than non-AI startups in the same industry and cohort, with 13% higher engineer shares, roughly 15% lower shares of entry-level workers and managers, flatter hierarchies, and comparable valuations. They argue embedding AI into products lets firms scale knowledge work without large teams.
  • Which Entrepreneurs Boost Productivity? — The authors find that a minority of entrepreneurs are "transformative" — generating disproportionate productivity gains, scoring high on IQ tests, being well-educated, and hiring research and development workers. They argue that economies with more such entrepreneurs and R&D workers sustain higher long-run productivity growth.
  • Innovation/Research
  • Growth
    • US Business Dynamism
  • Productivity
    • Startups
  • Workforce
    • Unemployment/Participation

Chart of the Day: Small Models are Closing the Gap to Frontier AI

AI Summary. Achieving 60% on MMLU, a test of academic reasoning, once required a 540bn-parameter model but was reached by a 3.8bn-parameter model two years later, a 142x reduction in parameters. Models small enough to run on phones or sensors shift inference from data centers to devices, weakening the case for paying cloud providers per token.

Paul Kedrosky Applied Complexity
Date Posted:
June 3, 2026
Is Database:
Database

Small models have increasingly closed the performance gap with frontier models. To Kedrosky, this implies “edge compute becomes real,” which in turn, at the margin, means reduced relative demand for data center-based versus local compute.

Are small AI models closing the gap with frontier models?

Core argument: Reaching 60% on MMLU, a test of academic reasoning, required a 540bn-parameter model in 2022 but only a 3.8bn-parameter model by 2024, a 142-fold parameter reduction in two years.

A quiet story in AI isn't that large models keep getting better, it's that small ones are catching up quickly. In 2022, reaching 60% on MMLU, a test of academic reasoning, required a 540-billion-parameter model. By 2024, a 3.8-billion-parameter model hit the same threshold — a 142-fold parameter reduction in two years. The direction is clear, and the implications are wide-ranging. Edge compute becomes real. Models that fit on a phone or a sensor don't need a data center. Inference shifts from hyperscaler to device, which is a different supply chain and a different competitive moat than anyone has built for. The data center API business model has a half-life. If a fine-tuned 7B model running locally can handle 80% of enterprise use cases, the case for paying per-token to a cloud provider shrinks fast. The most exposed providers are those whose moats are model quality rather than distribution, i.e., the capex-heavy frontier models. Compute becomes less scarce.

Takeaways by Macro Roundup® AI

  1. Reaching 60% on MMLU, a test of academic reasoning, required a 540bn-parameter model in 2022 but only a 3.8bn-parameter model by 2024, a 142-fold parameter reduction in two years.
  2. Models small enough to run on a phone or sensor shift inference from hyperscalers to devices, which involves a different supply chain and competitive moat than providers have built for.
  3. The case for paying per-token to a cloud provider shrinks if a fine-tuned 7B model running locally can handle 80% of enterprise use cases, leaving capex-heavy frontier providers whose moats rest on model quality rather than distribution most exposed.

Related Articles:

  • Commoditization, Orchestration, and the New AI Stack — The author argues that as incremental gains from larger AI training runs shrink, the return on additional computing power falls, weakening the case for the training spending that dominates data center demand. Survivors of model commoditization will be those owning the orchestration layer or the applications where outcomes are priced.
  • The Geopolitics of AI: Decoding the New Global Operating System — The performance gap between the top American and Chinese AI models narrowed from 9.3% to 1.7%, with Chinese firms relying on cheaper open-source models—DeepSeek prices usage 20-40x below OpenAI's ChatGPT. This price advantage could win adoption in developing countries and build dependency on Chinese AI ecosystems.
  • Misanthropic: On Mythos, Bad Human Behaviors and Systems Vulnerabilities — Anthropic reports that its Mythos model has detected thousands of high-severity cyber vulnerabilities, including ones found by chaining multiple obscure software weaknesses, and that these hacking skills are emergent rather than a designed capability. Mythos completed a corporate network attack simulation estimated to take an expert over 10 hours, which no other frontier model had solved.
  • Innovation/Research
  • GDP
    • Financial Markets
  • Productivity
    • Investment
© Copyright 2026 Coherent Research Institute · All Rights Reserved · Privacy · Terms