“…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
“…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
“…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
“…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
“…a must-read for serious students of economic policy.” - Glenn Hubbard, Dean, Columbia Business School, and former Chairman of the Council of Economic Advisers
“…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
“Unintended Consequences is full of substance, it is one of the must-read books of the year, and once I finish it I will be giving it a second read through right away.” - Tyler Cowen, Professor, George Mason University
“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
“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
“…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
“…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
AI Summary.Wall Street banks are pressing large law firms to cut fees as artificial intelligence speeds up legal tasks, with Morgan Stanley's general counsel calling the billable-hour compensation model unstable and shifting most outside legal work to competitive bidding and fixed fees. He argues firms could stay as profitable by using AI to handle more matters at lower cost.
Kaye Wiggins and Joshua FranklinFinancial Times
Date Posted:
AI is destabilizing the hourly-billing model of large law firms. Major banks are moving toward competitive bidding and fixed fees, expecting AI savings to lower legal costs, while firms preserve profits by handling more matters with fewer hours.
Is AI pushing law firms away from billable-hour pricing?
Core argument: Morgan Stanley general counsel Eric Grossman told the FT that law firm pay models built on associates billing long hours are "extraordinarily unstable," and that completing tasks more quickly could mark "a fundamental altering of the revenue foundation.".
Top lawyers have “for a long time been compensated on the foundation of [associates billing for long hours],” Eric Grossman, Morgan Stanley’s general counsel, told the FT. “Their compensation model is now extraordinarily unstable.” The ability to complete tasks more quickly could mark “a fundamental altering of the revenue foundation.” Grossman said the bank was willing to continue to pay large sums for the judgment and talent of the best lawyers, but that by the end of this year most external legal work would be tendered through competitive bidding processes and paid for using alternative arrangements such as fixed fees. That should cost the bank less, he said, but law firms could remain as profitable as before if they use AI to work on more matters and reduce costs.
Takeaways by Macro Roundup® AI
Morgan Stanley general counsel Eric Grossman told the FT that law firm pay models built on associates billing long hours are "extraordinarily unstable," and that completing tasks more quickly could mark "a fundamental altering of the revenue foundation.".
Grossman said Morgan Stanley would tender most external legal work through competitive bidding by the end of this year and pay via alternative arrangements such as fixed fees, which he said should cost the bank less.
Grossman said law firms could remain as profitable as before by using AI to work on more matters and reduce costs, and that Morgan Stanley remains willing to pay large sums for the judgment and talent of the best lawyers.
Related Articles:
AI and Productivity— The recent pickup in US labor productivity appears driven by higher utilization of existing capital—running factories, server racks, and hotel rooms harder—rather than by firm-level efficiency gains from AI. Adoption remains too thin and the cross-sectional evidence too weak to conclude that AI-driven productivity gains are already widespread.
US Focus: The Effect Of Soaring Profits— US nominal pre-tax corporate profits in the national accounts grew 23% over the past year, a pace not seen outside post-recession spikes since the mid-2000s. Wider margins were the key driver, as domestic profit growth far exceeded the 8% rise in corporate value added, pushing margins near all-time highs while the labor share hit new lows.
Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools— The authors estimate that AI coding tools raise developers' task-level productivity by about 40% with autocomplete, 140% including synchronous agents, and 180% with asynchronous agents, with larger effects for less active developers. Gains attenuate up the production chain: synchronous agents raise lines of code 741% and pull requests 65% but releases only 20%.
AI Summary.An analysis of millions of online job postings finds that openings fell for occupations whose tasks can be automated by generative artificial intelligence, with more-exposed firms advertising fewer such positions as a share of their total postings. Follow-up analyses suggest this shift in labor demand has negatively affected labor market outcomes for recent college graduates.
Samuel Dodini and Tucker SmithFederal Reserve Bank of Dallas
Date Posted:
A FRBD analysis of Lightcast (formerly Burning Glass) job postings data suggests AI-driven automation reduced job postings in Texas by ~1.8% in 2024 and ~2.6% in 2025, with recent college graduates most impacted.
Is AI automation reducing job openings for exposed occupations?
Core argument: An analysis of millions of online job postings finds that after ChatGPT's release in late 2022, Texas job openings fell for occupations whose tasks are automatable by generative AI.
A new analysis, using data from millions of online job postings, shows GenAI is reshaping labor demand in Texas. After the release of ChatGPT in late 2022, job openings fell for occupations whose tasks are automatable by GenAI. Indeed, more-exposed firms posted fewer automatable positions relative to their total ads after the release of ChatGPT. Follow-up analyses suggest this shift in demand has negatively affected the labor market outcomes of recent college graduates from Texas universities.
Takeaways by Macro Roundup® AI
An analysis of millions of online job postings finds that after ChatGPT's release in late 2022, Texas job openings fell for occupations whose tasks are automatable by generative AI.
Firms more exposed to generative AI posted fewer automatable positions as a share of their total job ads following ChatGPT's release, according to the analysis.
Follow-up analyses suggest the shift in labor demand has negatively affected labor market outcomes for recent graduates of Texas universities.
Related Articles:
Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence— Employment for workers aged 22–25 in the occupations most exposed to artificial intelligence has declined roughly 18pp relative to less-exposed peers, with no comparable decline in other age groups. The authors find this operates mainly through reduced hiring, concentrated where AI substitutes for human tasks rather than complements them.
You’re (Not) Hired: Artificial Intelligence and Early Career Hiring In The Quarterly Workforce Indicators— Employment among workers aged 22–24 in the most artificial-intelligence-exposed quintile of industry-state employers fell 15.2%, while other exposure quintiles were flat or posted modest gains. The decline is not generally observed among older workers; the 25-34 group's drop was less than one-third as large in percentage terms.
The Impact of AI on the U.S. Labor Market— Using observed AI usage data across 321 U.S. occupations, the authors estimate that highly exposed occupations saw a 6.7% decline in real wage growth with no detectable employment effects. The estimated declines are concentrated among the lowest earners, with service workers down 24.3% and the bottom wage quartile down 10.7%, while top earners show no significant effect.
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-EspínSocial Science Research Network
Date Posted:
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
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.
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.
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.
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 KedroskyApplied Complexity
Date Posted:
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
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.
The estimated change in slope of the AI capabilities trend carries a confidence interval of -8.4 to +23.4 points per year.
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.
AI Summary.Chinese patent data indicate that critical technology patenting is dominated by private firms and public universities rather than state-owned enterprises, with universities playing a larger role and large firms a smaller share than in the United States. Fewer than 10% of these patents involve inventors with U.S. training or employment.
Josh Lerner, Namrata Narain, Dimitris Papanikolaou, Amit Seru, et al.National Bureau of Economic Research
Date Posted:
China’s critical-technology innovation is produced chiefly by private firms and universities, not directly by state-owned enterprises or the PLA.
Who drives China's critical technology patenting?
Core argument: Fewer than 10% of Chinese critical technology patents involve inventors with U.S. training or employment, and the U.S. patent system relies substantially more on inventors with foreign backgrounds than the Chinese system does.
Patents, while an imperfect measure, have long been used to understand innovation in Western nations. In this paper, we use Chinese domestic awards to shed light on the Chinese innovative ecosystem. We suggest that many preconceptions about Chinese innovation are at odds with what is revealed by the patent data. Chinese critical technology patenting is dominated by private firms and public universities rather than state-owned enterprises, with universities playing a far larger role than in the United States. Also distinct from the United States is the low overall share of large firms. Direct foreign inventor linkages remain limited: fewer than ten percent of Chinese critical technology patents involve inventors with U.S. training or employment, while the U.S. patent system itself exhibits substantially greater reliance on inventors with foreign backgrounds than the Chinese system. Finally, the increase in Chinese critical technology patenting has not come at the cost of quality. Chinese patents have converged towards or even exceeded U.S. patents using several textual measures in most critical technologies, despite the massive increase in volume.
Takeaways by Macro Roundup® AI
Fewer than 10% of Chinese critical technology patents involve inventors with U.S. training or employment, and the U.S. patent system relies substantially more on inventors with foreign backgrounds than the Chinese system does.
Chinese critical technology patenting is dominated by private firms and public universities rather than state-owned enterprises, with universities playing a far larger role and large firms a smaller share than in the United States.
The rise in Chinese critical technology patenting has not come at the cost of quality, as Chinese patents have converged toward or exceeded U.S. patents on several textual measures in most critical technologies.
Related Articles:
China and the Future Of Science— China graduates about 5x as many medical and biomedical students as the US each year, 7x as many engineers, and nearly double the number of science and engineering PhDs, while one-sixth to one-fifth of US science and engineering graduates are themselves Chinese. One study found returning Chinese scientists lead-author 2.5x more papers than peers who stay in the US.
The Growing Self-Reliance of Chinese Innovation— Linking patents to scientific papers through shared phrases, the authors find China's innovation has grown markedly more self-reliant across nearly all technology areas the U.S. treats as strategically important. In an aggregate grouping of these areas, the share of patent phrases traced to Chinese-produced articles rose from 2% to 24%, while the U.S.-produced share fell from 31% to 18%.
China’s Next-Generation Industrial Policy— China's export growth is concentrated in upstream, higher-tech goods: exports of intermediate inputs rose 26% and capital goods 32% between 2021 and 2024, while consumption goods exports were broadly flat. This shift embeds China deeper in global production systems and aligns with its long-running import substitution strategy of localizing entire value chains.
AI Summary.Demand for frontier artificial intelligence models exceeds supply because computing capacity is scarce, allowing chip suppliers and their customers to resell compute at high margins. The author argues alarm over Chinese models is overblown, expecting the inference market to grow faster than training costs so labs can thrive at lower prices.
Ben ThompsonStratechery
Date Posted:
US AI labs look pricier than Chinese rivals right now — but that’s due to a transitory compute shortage, not a gap in the underlying cost structure. Once US compute scales up, the apparent gap will disappear.
Can AI labs thrive at lower prices as inference demand grows?
Core argument: Demand for frontier AI models exceeds supply because compute is scarce, allowing Nvidia to earn very large margins while its customers, such as SpaceXAI, resell compute at high margins to buyers like Anthropic.
Right now demand exceeds supply for frontier models, and supply is limited by a lack of compute. This computer shortage doesn't just mean that a compute supplier like Nvidia makes very large margins, but also that Nvidia's customers, like SpaceXAI, can turn around and resell compute at high margins as well to a company like Anthropic. Anthropic, meanwhile, can pay the markup because they can sell tokens with a higher markup still. I think the reaction to Kimi and Chinese models generally is pretty over-blown, at least from an economic perspective. As long as training consumed more GPUs than inference, it was critical to maximize inference revenue to help fund the next training run, which meant charging very high prices for inference. Going forward, however, I expect the inference market to grow much faster than training costs, which means they really can make it up in volume... frontier labs should have more confidence that they can not just survive but thrive with lower prices (once they have sufficient compute).
Takeaways by Macro Roundup® AI
Demand for frontier AI models exceeds supply because compute is scarce, allowing Nvidia to earn very large margins while its customers, such as SpaceXAI, resell compute at high margins to buyers like Anthropic.
Anthropic can absorb compute markups because it sells tokens at a still higher markup.
The reaction to Kimi and Chinese models is overblown from an economic perspective, since the inference market is expected to grow much faster than training costs, letting frontier labs thrive at lower prices once they have sufficient compute.
Related Articles:
China’s Moonshot Unveils AI Model That Narrows Gap With US Firms— Chinese startup Moonshot priced its latest AI model at roughly the same level as Anthropic's Sonnet, a premium over other Chinese models, and said it has 2.8tn parameters and a 1m token context window. Artificial Analysis ranked the model ahead of Anthropic's Opus 4.8 on some frontier benchmarks, the first Chinese open-weight model to do so.
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.
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.
AI Summary.Technology-driven investment booms have historically ended in sharp corrections with wider economic fallout, and the current artificial intelligence build-out is on track to outgrow every previous episode within three years measured from its pre-boom trough. The expansion is increasingly funded by borrowing through bonds, special purpose vehicles and non-bank lenders, alongside circular financing arrangements that create interconnectedness among firms.
Phurichai RungcharoenkitkulBank of International Settlements
Date Posted:
A BIS analysis argues the AI buildout is outpacing prior infrastructure booms and warns, “The larger the boom, the deeper the eventual bust.” Their model suggests that investment is 50% above the efficient level.
Is the AI build-out outpacing past technology investment booms?
Core argument: Firms expanding AI capital expenditure have turned increasingly to borrowing through bond issuance, special purpose vehicles, and non-bank lenders, and such leverage aggravates the fragility of rapid investment ramp-ups.
Technological breakthroughs are typically accompanied by investment booms and buoyant macroeconomic activity. Exuberance about the promise of new technologies intensifies competition among firms eager to capture a share of the revenues. The race to get ahead can result in excessive investment that makes the boom unsustainable and prone to disruption. This fragility is further aggravated by the leverage that accompanies the rapid ramp-up of investment. This boom-bust pattern recurs across history, from the US canal mania in the 1830s and the British railway mania in the 1840s to the roaring 20s and the dotcom boom in the late 90s. These episodes all ended in sharp corrections with wider economic fallout. Relative to its pre-boom trough, the current build-out is on track to outgrow every previous episode only three years. The rapid capex expansion has led some to resort increasingly to borrowing, not only through bond issuance but also through special purpose vehicles and non-bank lenders. Some firms have also engaged in so-called circular financing, in which a firm engaging in the build-out takes an equity position in an AI lab in exchange for the lab’s commitment to purchase future compute. Such arrangements have enabled greater deployment of capital, but they also create interconnectedness, shown in panel (b) of Figure 1.
Takeaways by Macro Roundup® AI
Firms expanding AI capital expenditure have turned increasingly to borrowing through bond issuance, special purpose vehicles, and non-bank lenders, and such leverage aggravates the fragility of rapid investment ramp-ups.
Circular financing, in which a firm engaged in the build-out takes an equity position in an AI lab in exchange for the lab's commitment to purchase future compute, has enabled greater capital deployment while creating interconnectedness.
Related Articles:
AI Set to be Largest CapEx Cycle Ever … and Soon Majority Externally Financed— Hyperscaler and data-center investment is tracking toward $1.5tn annually by 2030, which in real terms would exceed the 2005-06 housing boom, the 1990s fiber peak, and the 1882 railroad peak combined. External debt and equity are projected to fund close to 90% of the buildout, making it dependent on credit markets.
Have We Entered an Era of High Productivity Growth?— Labor productivity has grown solidly while the San Francisco Federal Reserve's total factor productivity index, which adjusts for the equipment and technology workers use, has grown modestly, suggesting gains so far reflect better tools rather than broader advances. A regime-switching model puts the probability of a high-productivity regime at 57% using labor productivity and 21% using total factor productivity.
The Macro Implications of the AI Capex Boom— Artificial intelligence capital spending is estimated to add roughly 1.4 percentage points to US growth in one year and about 1.5 points the next, comparable to business investment's contribution during the tech bubble. A material part of that estimate reflects profits at Nvidia and a few other chipmakers that are unlikely to be recycled back into the economy.
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 MsikaBloomberg
Date Posted:
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
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.
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.
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.
