Friday, June 13, 2025

This week's interesting finds

This week in charts

S&P 500 Index valuations vs. their historical average

U.S. vs. global equity price returns

U.S. stock market value relative to GDP

U.S. value stocks vs. growth stocks

Japanese share buybacks

EV/EBITDA since 2005

European equities cash deployment

Corporate bond spreads

U.S. bond and equity correlation

U.S. Aggregate bond index duration vs. High Yield corporate bond duration

Automotive sales by region

As Companies Abandon Climate Pledges, Is There a Silver Lining?

Peter Ford has seen the promise—and the pitfalls—of corporate climate pledges up close. The 40-year-old Briton recently spent five years at Hennes & Mauritz AB trying to cut emissions from the Swedish fashion giant’s vast supply chain, from Cambodian sewing lines to Vietnamese dye houses. He met with hundreds of suppliers, pushed for energy efficiency upgrades and urged the elimination of coal boilers. And to its credit, H&M invested about $200 million a year in these efforts and recently reported a 24% cut in its supply chain emissions.

But Ford isn’t celebrating. “As an industry, it’s not working out yet,” he says bluntly. Apparel emissions are still growing—and could expand an additional 30% this decade, according to McKinsey & Co. While a few brands are doing a lot of work, Ford says, most of the industry “would much rather sit there and wait for things to happen.” That mismatch isn’t unique to fashion. From airlines to banks to retailers, the story is the same: Over the past few decades, more than 4,000 companies have made big climate pledges, but results are scant, and emissions continue to rise.

Worse, we’re now seeing a retreat. In the past year companies around the world have been canceling their climate commitments, some only a few years old. BP Plc is pulling back on renewables and drilling more oil. Coca-Cola Co. and PepsiCo Inc. abandoned or weakened promises they made in 2021 to slash their use of new plastics. Big banks such as Wells Fargo & Co. and HSBC Holdings Plc walked back various plans to reduce their emissions. Walmart Inc. admits it’s behind on its climate targets, while FedEx Corp. says it will likely miss its goal to go electric on half of its delivery truck purchases by 2025. This corporate retrenchment has been particularly acute in the US, where the Trump administration has been busy rolling back climate regulations and withdrawing from international treaties such as the Paris Agreement.

It’s a troubling trend for anyone who prefers life on a hospitable planet. We’re currently on pace to add about 3C (5.5F) of warming this century, which is expected to throw the planet into turmoil with diminished food supplies, wiped-out marine life, brutal heat waves and crippling droughts. But even with those grim prospects, some experts argue that this corporate backpedaling might come with a silver lining. That’s because it could force investors, lawmakers, academics and the broader public to reckon with the fact that voluntary corporate action was never going to stave off climate disaster. This could bring sharper attention to corporate political activities, where even self-proclaimed responsible businesses obstruct regulations needed to phase out fossil fuels and boost clean alternatives.

For years companies have suggested their pursuit of sustainability would naturally follow their quest for profits. Walmart framed its eco-pivot in 2005 as a way to reduce waste, cut costs and promote goodwill. PepsiCo’s then-chief executive officer, Indra Nooyi, put it more elegantly a few years later, when she said using less water and energy isn’t corporate benevolence but rather a way to lower your bills and fatten your margins.

There’s an element of truth to this: Some climate-friendly endeavors, including swapping out lightbulbs and putting up solar panels, quickly pay for themselves and enhance the bottom line. But most projects required to achieve deep decarbonization, such as decommissioning coal boilers or using cleaner fuels, cost gobs of money, and no one knows when or if they’ll ever be cheaper than their dirtier alternatives. As long as these measures remain voluntary, they won’t happen at anywhere near the scale or pace that’s needed. And companies will be free to renege on their promises.

Unfortunately, most companies have talked a big game on climate while working to block or water down the very policies that could drive real progress. Take the US airline industry. Despite skyrocketing emissions, most carriers have vowed to eliminate their planet-warming pollution by 2050. A key part of the plan is to use vastly more sustainable aviation fuel. But this currently accounts for about 0.3% of their total fuel use. Airlines are quick to point out that cleaner fuels cost about two or three times more than conventional jet fuel, and they can’t boost these purchases without putting themselves at a disadvantage to others that keep using cheaper, dirtier fuels.

This, of course, highlights the need for regulation to make sure first movers aren’t punished in the market. In Europe, regulators have stepped in, mandating 2% cleaner fuel now and 6% by 2030. But US carriers vigorously oppose rules like these, including a recent California proposal that would have regulated jet fuel for flights in the state. After airlines questioned the state’s authority to set such rules, California backed off and instead joined the companies to cheer a toothless “voluntary and nonbinding” agreement to increase the use of cleaner fuels.

When asked about these glaring misalignments, companies frequently argue that they rely on trade groups to advocate on a wide range of issues, such as taxation and trade, and that they don’t always agree on every position. This claim doesn’t pass muster with Bill Weihl, who says he used to make the same argument back when he ran sustainability efforts for Facebook. “It’s a weak excuse,” says Weihl, who has since launched ClimateVoice, a nonprofit that pushes companies to support climate policies. “They’re choosing to kick the can down the road so they can get lower taxes and deregulation.”

When it comes to policy positions, who’s putting their money where their mouth is? Advocates point to packaged-goods maker Unilever Plc as one of the rare companies moving in the right direction by regularly examining the climate work of trade groups to which it belongs. In its most recent review, from April, Unilever found that eight of 26 groups were misaligned with its views on climate policy—including the Confederation of Indian Industry, which has pushed for lower petrol taxes and more energy from gasified coal. Importantly, the report also spells out the steps Unilever plans to take to address this misalignment, including potentially leaving groups that continue to obstruct climate action. Any company claiming to be serious about climate should follow Unilever’s lead with this kind of rigorous examination and public reporting.

Clearly regulations are the most effective way to shift this mindset and force companies to pour money into much-needed climate projects. Businesses, however, have gone mute as President Donald Trump wades into their industries to annihilate various climate rules, which will pour billions of extra tons of planet-warming gases into the atmosphere. But there are other ways to pressure a company into speaking up and doing the right thing, including rallying its often large numbers of employees who feel strongly about leaving a livable planet for their children.


This week’s fun find

Mysteriously Perfect Sphere Spotted in Space by Astronomers

Our Milky Way galaxy is home to some extremely weird things, but a new discovery has astronomers truly baffled.

In data collected by a powerful radio telescope, astronomers have found what appears to be a perfectly spherical bubble. We know more or less what it is – it's the ball of expanding material ejected by an exploding star, a supernova remnant – but how it came to be is more of a puzzle.




Friday, June 6, 2025

This week's interesting finds

This week in charts

Foreign investment in U.S. fixed income assets

Momentum crowding still at extremes

Sector contribution to 2025 EPS growth estimates

Forward price-to-earnings ratios by geography

MSCI U.S. price-to-earnings decline during past recessions

MSCI Japan Index price-to-book relative to MSCI World Index

Share of domestic equity ownership and valuations

Share of U.S. firms using AI by sector

Housing vs. rental affordability

Big investors shift away from US markets

The US president’s erratic trade policy has shaken global markets in recent months, sparking a sharp sell-off in the US dollar and leaving Wall Street stocks lagging far behind European rivals this year.

Trump’s landmark tax bill, which is forecast to add $2.4tn to Washington’s debt over the next decade, has also increased pressure on US Treasuries.

The move away from US assets has pushed up European markets at the expense of their US counterparts and has been signalled by surveys of big institutional investors’ allocation decisions. A poll of fund managers published by Bank of America last month showed the biggest underweight in the US dollar in nearly two decades.

As institutional investors review the extent of their holdings in the US, Caisse de dépôt et placement du Québec, Canada’s second-largest pension fund, said recently it would reduce its exposure to the country — currently 40 per cent of its portfolio. It plans to increase investment in the UK, France and Germany. 

US stocks have recouped the losses that followed Trump’s announcement of the duties on April 2. But the S&P 500 remains less than 2 per cent up this year, compared with 9 per cent for the Stoxx Europe 600 index.

The dollar is close to a three-year low — down 9 per cent this year — even though Trump has retreated on many of the tariffs he initially announced.

Investors say that the global dominance of the US economy and the depth of its capital markets mean it will remain the premier destination for global investment.

However, many are questioning whether more than a decade and a half of inflows and outperformance — which pushed the US share of global equity market value to around two-thirds by the start of this year — is headed into reverse.

Some investors question whether smaller, more fragmented markets in Europe and Asia offer a meaningful alternative.


This week’s fun find

World’s First Mass-Produced Flying Car Prototype Unveiled

Eager teens reaching driving age in the next few years may be able to take their inaugural spin in a car … in the sky. The world’s first mass-produced flying automobile prototype has been unveiled, and we’re ready for a ride.

Created by Slovakia-based company Klein Vision, the AirCar production prototype made its public debut May 8, after making its insider debut at the 2025 Living Legends of Aviation Awards Ceremony in Beverly Hills late last month. At the event, Morgan Freeman and John Travolta presented the car’s inventor, Stefan Klein, with a Special Recognition Award for Engineering Excellence.

Friday, May 30, 2025

This week's interesting finds

17th annual Cymbria Investor Day

This year the Investment Team talked about the importance of identifying change in the businesses they invest in. The discussion includes analysis of past opportunities and the ways they believe they can continue to find new ones in the future. 

Click here to watch the video. 


This week in charts

U.S. data center gas demand

Historical asset bubbles

U.S. government & corporate debt

Days working from home, per week, by country

S&P 500 Index median stock short interest

U.S. equity index price-to-earnings valuations

S&P 500 Index forward price-to-earnings valuations

S&P 500 Index defensive sector allocation

U.S. value stocks vs. international value stocks

U.S. manufacturing employment

UK confirms powers to force pension funds to back British assets

Rachel Reeves has confirmed she will create a “backstop” power to force large pension funds to back British assets, as she vowed to unleash more than £50bn of investment in domestic infrastructure, housing and fast-growing businesses.

The chancellor hopes creating pension “megafunds” with more than £25bn in assets, coupled with a voluntary accord with industry to boost allocations to private assets, will reverse long-term falls in investment in the UK.

It is the first time the Treasury has publicly confirmed it will legislate to create a backstop power to mandate pension funds on their investment strategy — a move condemned by the Conservatives.

Reeves’ allies say she believes the power will not be needed and that reforms in the new pensions legislation — plus a “Mansion House accord” with the industry — would deliver the desired results.

At the heart of the reforms is a requirement for all multi-employer defined contribution pension schemes and local government pension scheme pools to operate at a “megafund level”, similar to those in Australia and Canada.

The Treasury argues that the megafunds, managing at least £25bn in assets, will drive more than £50bn of investment in big UK infrastructure and other British assets. This will then drive higher returns for savers, most of whom are covered by DC schemes.

However, after pushback from some industry participants to plans presented last autumn, the Treasury said schemes worth more than £10bn that are unable to reach the minimum size requirement by 2030 will be allowed to continue operating, as long as they can demonstrate a clear plan to reach £25bn by 2035.

The UK has about 60 multi-employer DC pension schemes, with combined assets forecast to reach £800bn in five years.

Meanwhile, under a Mansion House accord signed this month, 17 of the UK’s largest pension providers have pledged to invest at least 5 per cent of their default funds in private British assets by the end of the decade.

The Treasury said the Mansion House pact would release £25bn for UK investment by 2030, while a similar amount would come from local investment targets from local government pension schemes.

Reeves’ aim is to reverse a sharp decline in domestic investment from UK pensions funds, which the Treasury said had fallen from more than 50 per cent of DC assets in 2012 to roughly 20 per cent in 2023.

Blanc said this month that forcing funds to buy UK assets would not be “the right thing”, adding: “It’s like using a sledgehammer to crack a nut.”

Shadow chancellor Mel Stride said: “By pressing ahead with their plans to mandate pension-fund investments, Labour is crossing the Rubicon into directing the public’s savings. This is an extraordinary over-reach.”


This week’s fun find

Take a Nostalgic Dive Through a Visual Cassette Tape Archive

The cassette tape, invented in 1963, entered the market with a lukewarm reception as it competed with reel-to-reel and 8-track technologies. The suitability for recorded music along with its portability eventually put it on top of its competitors, and sound quality continued to improve in the 1970s. By the following decade, the cassette was a favorite among consumers, overtaking vinyl and continuing to dominate until the 1990s, when CDs superseded the technology.

Friday, May 23, 2025

This week's interesting finds

This week in charts

U.S. mobile traffic

Liquefied Natural Gas (LNG) demand

LNG imports by region

Electrification

U.S. 10-year Treasury yield since 1790

Reserve currencies

Hedge fund net flows into Industrial stocks

Growth in total employment - Magnificent 7

U.S. and Euro area banks bond holdings - Share of total assets

High-yield fund flows

Bond spreads

Average tariff rates

Researcher demographics

The Japanese government bond market is in trouble

The Japanese government bond market was already having a bit of a springtime nightmare, but a poor auction of 20-year debt earlier this week has sent long-end yields soaring to their highest levels ever.

The 40-year Japanese bond now yields nearly 3.7 per cent, up a full percentage point just since the beginning of April. That hurts. If the duration of the bond is, say, 20 years, investors have lost almost 20 per cent in just a few weeks. In a high grade government bond. That sort of thing just isn’t supposed to happen in the developed markets; bonds are supposed to be the safe asset class.

The Japanese yield curve is upward sloping — in fact, the steepest yield curve in the developed world. That means longer maturity bonds should have higher yields. And yet, the 35-year JGB now yields over 100 bps more than the 40-year one, simply because it is less liquid and “off-the-run”. That kind of glitchy pricing is pretty stark evidence of the evaporating demand for longer-term Japanese debt.

Japan’s fiscal problems are well known, of course. The country has had a massive debt burden for many years. Its credit rating is three notches below the US even after the recent US downgrade. Prime minister Shigeru Ishiba called Japan’s fiscal position “worse than Greece’s” — presumably referring to the Greece of the European sovereign debt crisis (Greece is in OK shape right now). None of this is new.

But several things have changed recently.

First is inflation. Headline CPI in Japan is now 3.6 percent, and has been above 2 per cent for over three years. That means that a 30-year bond yield at 3.15 per cent is still below inflation, or negative in real terms.

Second, Japanese insurers have virtually stopped purchasing very long-term bonds, to satisfy new economic solvency ratios introduced recently.

Third, long-term fiscal concerns are rising. The US is pressuring Japan to spend 3 per cent of GDP on defence (from the current 1.6 per cent). The ruling coalition is considering a supplementary budget, while the opposition is pushing for a consumption tax cut. No one seems to be talking about less spending and raising more tax revenues.

Perhaps the most important change is from the Bank of Japan. The central bank still owns a staggering 50 per cent-plus of the Japanese government bond market. But with deflation now in the rear view, the BoJ has started quantitative tightening — allowing its existing bond holdings to slowly hit the market. And as this one massive buyer has become a net seller, Japanese bond yields are normalising — aka rising.

Global bond markets are clearly paying attention. US Treasuries shrugged off the Moody’s downgrade, but since Tuesday the 30-year US Treasury yield has risen almost 20bp — the highest level since the 2008 crisis — to well over 5 per cent after dismal auction yesterday. As a result, US mortgage rates are back above 7 per cent.

There has been no data to speak of, and the tax bill that just passed the House contained no new surprises for the market. The main driver of the US bond move seems to be what is happening 7,000 miles away in Tokyo. In fact, the long end of pretty much every major bond market is getting caned in a broad-based duration crash. 

Japan bulls will point out that the last few weeks are not a repudiation of Japan Inc — and they’re right. No one is fleeing the yen, and the Nikkei is still up for the year in dollar terms. The economic impact is limited by the fact that most Japanese debt — mortgages, corporate debt, etc — is issued in shorter bonds.

But. . . Japanese policymakers are now at serious risk of losing control of the long end of the Japanese yield curve, absent imminent and forceful intervention. And if that happens, it’s bad news for everyone.


This week’s fun finds

Nothing brings people together more than great food and even better company culture. The moai hosted by TJ from the Trading team (with an assist from his better half Tiffany), was no exception. The sausages, schnitzel sandwiches and Detroit-style pizza was simple, satisfying and exactly what we needed to end the week. 

The world’s longest train journey is epic — but nobody’s ever taken it

The mountains of northern Laos are beautiful, but tough to negotiate. By car, it can easily take 15 hours to drive the 373 miles (600 km) of winding roads that separate the capital Vientiane from the town of Boten on the Chinese border.

Since December 2021, there’s a far straighter, much faster alternative: the brand-new high-speed Laos-China Railway (LCR) measures just 257 miles (414 km) between Boten and Vientiane, and fast trains cover that distance in three and a half hours.

The line is a marvel of engineering: It includes no fewer than 75 tunnels, which make up 47% of its total length, and 167 bridges, accounting for a further 15%. By all accounts, the views — outside of those tunnels — are spectacular. But the LCR is more than a scenic extension of China’s Belt and Road Initiative. For train enthusiasts around the globe, it is the final piece of a much grander puzzle — for this stretch is also part of the longest possible train journey in the world.

Friday, May 16, 2025

This week's interesting finds

This week in charts

Cumulative flows to U.S. equity funds

U.S. fund flows by market-cap

Small cap performance

Small cap vs. large cap 10-year performance

Nifty Fifty vs. Magnificent 7 vs. S&P 500 Index

Historical Dow Jones Industrial Average & Fed Funds Rate %

AI adoption rate by sector

U.S. consumer savings rate

Distribution of expenditures on imports

DeepSeek’s ‘Tech Madman’ Founder Is Threatening US Dominance in AI Race

DeepSeek cleared the fogged window through which Americans have viewed much of China’s AI scene: shrouded in mystery, easier to dismiss as an exaggerated specter but very likely more daunting than they’re willing to admit. Before the startup’s emergence, many US companies and policymakers held the comforting view that China still lagged significantly behind Silicon Valley, giving them time to prepare for eventual parity or to prevent China from ever getting there.


Where China sees innovation, many in the US continue to suspect malfeasance. An April report from a bipartisan House of Representatives committee alleged “significant” ties between DeepSeek and the Chinese government, concluding that the company unlawfully stole data from OpenAI and represented a “profound threat” to US national security. Dario Amodei, CEO of Anthropic, has called for more US export controls, contending in a 3,400-word blog post that DeepSeek must have smuggled significant quantities of Nvidia GPUs, including its state-of-the-art H100s. (Bloomberg News recently reported that US officials are probing whether DeepSeek circumvented export restrictions by purchasing prohibited chips through third parties in Singapore.)

The Chinese Embassy has rejected the House committee’s claims as “groundless.” Nvidia has said that DeepSeek’s chips were export-compliant and that more restrictions could benefit Chinese semiconductors. A spokesperson for the chipmaker says forcing DeepSeek to use more chips and services from China would “boost Huawei and foreign AI infrastructure providers.”

The company at the center of this debate continues to be something of an enigma. DeepSeek prides itself on open-sourcing its AI technology while not being open whatsoever about its inner workings or intentions. It reveals hyperspecific details of its research in public papers but won’t provide basic information about the general costs of building its AI, the current makeup of its GPUs or the origins of its data.

To further understand how the company works and how it fits into the country’s broader AI ambitions, Bloomberg Businessweek spoke with 11 former employees of Liang’s, along with more than three dozen analysts, venture capitalists and executives close to China’s AI industry.

The lack of a public presence has allowed critics such as Amodei and OpenAI head Sam Altman to fill the void with aspersions, which resonate with US audiences who are primed to see Chinese technology as a shadowy threat. But even those who remain wary of DeepSeek are being forced to grapple with the undeniable prowess of its AI. Dmitry Shevelenko, the chief business officer of Perplexity AI Inc., says not a single person at his company, which makes an AI-powered search product, has managed to communicate with any counterparts at DeepSeek. Nevertheless, Perplexity has embraced DeepSeek’s tech, hosting it only on servers in the US and Europe and post-training it to remove any datasets indicative of CCP censorship. Perplexity branded it R1 1776 (a reference to the year of the US’s founding), which Shevelenko describes as an homage to freedom. “We don’t know what DeepSeek’s true motivations are,” he says. “It’s a bit of a black box.”

DeepSeek had anticipated its AI might cause concerns abroad. In an overlooked virtual presentation at an Nvidia developer conference in March 2024, Deli Chen, a deep-learning researcher at DeepSeek, spoke of how values ought to be “decoupled” from LLMs and adapted to different societies. On one coldly logical slide, Chen showed a DeepSeek prototype for customizing the ethical standards built into chatbots being used by people of various backgrounds. With a quick tap of a button, developers could set the legality of issues including gambling, euthanasia, sex work, gun ownership, cannabis and surrogacy. “All they need to do is to select options that fit their needs, and then they will be able to enjoy a model service that is tailored specifically to their values,” Chen explained.

Finding such efficient workarounds was always the cultural norm at DeepSeek. Liang and his friends studied various technical fields at Zhejiang University in the mid-2000s—machine learning, signal processing, electronic engineering, etc.—and, apparently for kicks (and, you know, for cash), developed computer programs to trade stocks during the global financial crisis.

After graduating, Liang continued building quant-trading systems on his own, earning a small fortune before joining forces with several of his university friends in Hangzhou, where they launched what became known as High Flyer Quant in 2015.

As would be the case with DeepSeek, High-Flyer cultivated a sense of mystery—its first social media post referred to Liang only as “Mr. L”—while committing itself to a kind of lemme-prove-it transparency. Every Friday, High-Flyer would post charts of the performance of its 10 original funds on the Chinese super-app WeChat. Before making the weekly data available only to registered investors in the summer of 2016, the portfolio was seeing average annualized returns of 35%.

Billions of dollars eventually flowed into High-Flyer’s holdings, and its investment and research group increased to more than 100 employees. Liang started recruiting in earnest for an AI division in 2019, aiming to mine gargantuan datasets to spot undervalued stocks, tiny price fluctuations for high-frequency trading and macro trends that industry-specific investors were missing. By the beginning of the Covid-19 pandemic, he and his team had constructed a high-performance computing system of interconnected processors running in tandem, a setup known as a cluster. For this cluster, High-Flyer said it had acquired 1,000 Nvidia 2080Ti chips—commonly used by gamers and 3D artists—and an additional 100 Volta-series GPUs. (The Volta GPU, aka the V100, was Nvidia’s first AI-optimized processor.) Whereas High-Flyer’s previous, smaller computing architecture required two months to train a new economic analysis model, its new equipment needed less than four days to process the same workload.

It’s unclear how much of this infrastructure was ultimately intended for quant trading versus Liang’s expensive hobby. The next spring, about five months after OpenAI introduced ChatGPT, he spun out DeepSeek as an independent research lab. At separate offices in Hangzhou and Beijing, finance was no longer the focus. In an unsigned manifesto rife with platitudes, High-Flyer vowed to shun mediocrity and tackle the hardest challenges of the AI revolution. Its ultimate goal: artificial general intelligence.

Throughout 2023 the DeepSeek lab raced to build an AI code assistant, a general-knowledge chatbot and a text-to-3D-art generator. Liang brought over engineers from High-Flyer and recruited more from Microsoft Corp.’s Beijing office and leading Chinese tech companies and universities. Bo “Benjamin” Liu, who joined as a student researcher that September prior to starting a Ph.D., says Liang frequently gave interns crucial jobs that elsewhere would be assigned to senior employees. “Take me as an example: When I got to the company, no one was working on the RLHF infra”—the infrastructure needed to support an important technique known as reinforcement learning from human feedback—“so he just let me do it,” Liu says. “He will trust you to do the things no one has done before.” (That trust came with a secondary benefit to DeepSeek: It paid interns the equivalent of $140 per day with a $420 monthly housing subsidy, generous compensation in China but about a third of what interns make at AI companies in the US, and a tiny fraction of what full-time Silicon Valley engineers earn.)

Liang placed a huge and early bet on sparsity, a technique for training and running LLMs more efficiently by breaking them down into specialties, according to two ex-DeepSeek researchers. When you asked the original ChatGPT a question, its entire LLM brain would activate to determine the ideal answer, whether you asked for the sum of 2 + 2 or a pie recipe. A sparse model, by contrast, would make better use of resources by being partitioned into “experts,” with only the relevant ones being activated in response to any particular prompt.

More breakthroughs followed, each shared publicly and increasingly catching the attention of Chinese competitors. Then, in late 2024, DeepSeek released V3, a general-purpose AI model that was about 65% larger than Meta Platforms Inc.’s equivalent, which was then the biggest open-source LLM available. But it was a lengthy V3 research paper that really grabbed the attention of executives at Google, OpenAI and Microsoft, about a month before DeepSeek broke into the wider consciousness with its R1 reasoning model. One shocking statistic that leapt off the PDF: DeepSeek implied that V3’s overall development had cost a mere $5.6 million. It’s likely this sum referred only to the final training run—a data-refinement process that transforms a model’s previous prototypes into a complete product—but many people perceived it as an insanely low budget for the entire project. By comparison, cumulative training for the most advanced frontier models can run $100 million or more. Anthropic’s Amodei even predicted (before the rise of DeepSeek) that next-generation models will each cost anywhere from $10 billion to $100 billion to train.

DeepSeek exhibited its rapid progress because Liang saw the open-source ethos as integral to his philosophy. He believed that hiding proprietary techniques and charging for powerful models—the approach taken by top US labs including OpenAI and Google—prioritized short-term advantage over more durable success. Making his models entirely accessible to the public, and largely free, was the most efficient way for DeepSeek to accelerate adoption and get startups and researchers building on its tech. The hope was that this would create a flywheel of product consumption and feedback. As DeepSeek wrote in the announcement of its first publicized LLM almost two years ago, quoting the inventor of open-source operating system Linux: “Talk is cheap, show me the code.”

DeepSeek, however, has presented itself as no different than any hot startup—the product of “pure garage-energy,” it said in a February post on X. After all, it operates on the same Beijing campus as Google, not far from a Burger King and two Tim Hortons. Just because the broader AI industry didn’t pay much attention to DeepSeek until now doesn’t mean something shady is happening behind the scenes. “The AI world didn’t expect DeepSeek,” says Arnaud Barthelemy, a partner at VC firm Alpha Intelligence Capital, which has invested in OpenAI and SenseTime. “They should have.”

Barthelemy says the real lesson to take from DeepSeek is how effectively Chinese tech companies are turning the constraints they operate under into a strength. “There are plenty of smart minds in China who did a lot of smart innovation with much lower compute requirements,” he says.


This week’s fund find

Unwind with the Ancient Japanese Art of Kumiko, a Wood Joinery Technique

If you’re familiar with the Japanese art of wood joinery, you’ll likely find kumiko equally intriguing. The traditional craft emerged in the Asuka era between about 600 and 700 C.E. and similarly eschews nails in favor of perfectly cut pieces that notch into place. Intricate fields of florals and geometric shapes emerge, creating a decorative panel that typically covers windows or divides a room.