TheChek News Flow

Databricks on Thursday announced a new round of funding that values the company at $188 billion. The round was led by Coatue.

Databricks didn’t disclose exactly how much it raised; it said the money isn’t in its hands yet and that the round will close later this summer. (Other outlets have since reported the raise is roughly $3 billion.) While it’s unusual for a company to announce before it gets the money, a VC tells TechCrunch that the deal is solid, with so many firms wanting in that the company had no reason to keep its shiny new valuation a secret.

In fact, Databricks has been on a year-and-a-half fundraising tear as it successfully transitioned its image into an AI provider and not just a yesteryear SaaS sensation. Yesteryear being back in the BC times (Before ChatGPT).

Only five months ago, in February, Databricks closed a $5 billion Series L raise at a $134 billion valuation. Five months before that, in September 2025, it raised $1 billion at a $100 billion valuation. And roughly nine months before that, in December 2024, it raised what was a record-breaking round at the time of $10 billion at a $62 billion valuation.

Databricks has raised so many rounds over the years that this latest one became the subject of memes about running out of letters of the alphabet. “Turning on alerts for when we get a Series AA,” one person posted.

But its image reconstruction has been legit. Founded in 2013, it initially grew to success back in the big data era, with software that enabled enterprises to store enormous amounts of data in the cloud, yet produce speedy analytics.

Because it already sat on troves of enterprise data, Databricks was then well-positioned to respond as companies started wanting AI with the same security and governance they expect from traditional enterprise software.

The company began rolling out one AI product after another, like Lakebase, its database built for AI agents, and Unity, its AI gateway, along with a “meta-harness” called Omnigent that manages multiple agents.

Databricks also increasingly became known as one of the big examples of enterprises adopting more affordable Chinese-based open-weight models (models whose underlying code is published for anyone to use and modify) for cost control, one of the big trends of 2026. It is a particular champion of Z.ai’s GLM 5.2 as a model for coding.

Last week Databricks CEO Ali Ghodsi shared the results of some internal benchmarking done to manage his own AI costs for his 3,000 software engineers.

The company compared AI models on the actual tasks its programmers do. Not surprisingly, in the blog post revealing the results, Databricks shared that “open models, and GLM 5.2 in particular, are now able to handle even the highest level of task difficulty” in coding, and at a total lower cost than proprietary models from Anthropic and OpenAI.

But it did surprise people by finding that the choice of harness — the agentic coding tool, like Codex or Claude Code, that wraps around a model and manages its context and instructions — equally impacted costs. It found that open-source harness Pi to be one of the best at managing context surrounding each prompt, and therefore one of the lowest costs choices without sacrificing quality.

“The lesson here isn’t that one harness is always cheaper or that native harnesses are worse,” the post declared. “Instead, model choice is only one piece of the puzzle.”

All of this has added to Databricks image as an AI company, even if it wasn’t founded as an AI lab. This, in turn, has granted it the AI-halo for raising money and leaping its valuation. As we previously reported, the AI effect is so strong these days, that even sandwich shop Jersey Mike’s mentioned AI 22 times in its S-1 documents.

AI Chip Startup Etched Is in Talks for $20 Billion Valuation

Startup is also raising capital at $10 billion valuation in separate round led by Sequoia

Etched, a startup developing AI chips in competition with Nvidia NVDA -2.21%decrease; down pointing triangle, is set to quadruple its valuation to about $20 billion in a new funding round led by an existing investor, the firm Jane Street, according to people familiar with the matter.

Benefiting from a frenzied market for venture-capital investments in artificial-intelligence startups, the San Jose, Calif.-based company is also raising capital at a $10 billion valuation in a separate round led by Sequoia Capital, the people said. The financings haven’t closed, and the terms of the deals could still change.

Such back-to-back financings have become more commonplace in Silicon Valley in the midst of the AI boom. Startups are frequently selling stakes at one valuation and then quickly raising more capital at far higher prices—a reflection of the bargaining power companies have over investors scrambling to invest in the leading AI players.

Etched has said it is building a chip for running AI models, also known as inference. According to the firm’s website, it is testing its initial chip design and working to validate its first product to fulfill $1 billion in demand from customers.

While Nvidia is the leading vendor of AI chips, a wave of startups is trying to challenge its dominance, buoyed by the success of Cerebras Systems and Groq. Many of the new startups are developing chips tailored for inference, including the U.K. startup Fractile and SambaNova. Nvidia’s dominance is rooted in the capability of its graphics processing units, or GPUs, at training AI models.

Etched was founded in 2022 by Gavin Uberti, Chris Zhu and Robert Wachen, who are Harvard dropouts. Earlier investors in the company include the investment firm Stripes, Peter Thiel, Ribbit Capital and Primary Venture Partners.

Strikes on more sensitive targets across the Persian Gulf risk setting off a spiral as neither side backs down

Fighting between the U.S. and Iran has expanded, with the American military striking a broader range of targets and moving jet fighters to the region.

The U.S. has conducted seven consecutive days of strikes, hitting bridges and other targets in Iran’s interior in an effort to stop attacks on Gulf shipping.

Iran has responded by broadening the geographical reach of its attacks, including strikes on Kuwait, Qatar, Oman and shipping vessels.

Fighting between the U.S. and Iran has expanded, with the American military striking a broader range of targets and moving jet fighters into the Middle East while Tehran launches attacks across the Persian Gulf.

The fighting is focused on control of the Strait of Hormuz but this round raises the risk of a return to a larger war. The U.S. is hitting bridges and other targets in Iran’s interior to increase pressure across Iran to force an end to attacks on Gulf shipping, after a deal to open the waterway collapsed last week. Iran has responded with broader and deadlier attacks.

A return to a wider conflict would put upward pressure on oil prices—which have already risen more than 10% this week—and weigh on the global economy, risks President Trump has made clear concern him. Iran also faces a massive rebuilding challenge and a population deeply unhappy with the government.

So both sides have incentives to avoid a major flare up. But each is pushing for leverage, raising fear of an escalatory spiral that gets out of hand, said Saeid Golkar, an expert on Iran’s security services who teaches at the University of Tennessee at Chattanooga.

“This escalation is rapidly intensifying and getting out of control,” he said. “There is a risk we will go back to a total war even if neither side wants it.”

The strait was supposed to be opened under a memorandum of understanding Trump signed with Iran a month ago. The deal fell apart amid new Iranian attacks on shipping aimed at shutting down a U.S. effort to shepherd traffic through the strait along the coast of Oman.

Iran believes the agreement gives it the right to manage traffic through the strait and wants ships to pass through a northern route along the Iranian coast.

U.S. officials said earlier this week that Trump was leaning toward expanding U.S. military operations to break the diplomatic logjam. The U.S. was moving jet fighters back to the Middle East from Europe, according to flight tracking data and a person familiar with the matter.

More than 2,000 Marines from the 11th Marine Expeditionary Unit are also operating in the region after spending time in the Pacific. The military released photos of the Marines boarding and searching a commercial vessel in the Gulf of Oman as the U.S. ramps up enforcement of its blockade of Iranian ports.

The U.S. has now carried out seven consecutive days of strikes, the biggest escalation since the preliminary deal was signed in June. Like other days, U.S. Central Command said Friday afternoon, Eastern time, that the new round of strikes are “designed to continue degrading Iranian military capabilities.”

Recent targets have included multiple bridges in an effort to cut off supply routes to the port and naval base at Bandar Abbas on the Strait of Hormuz. The port normally handles 90% of the country’s container traffic. Iran also uses the facilities to attack ships, The Wall Street Journal has reported, citing a senior U.S. official.

Several attacks on bridges were reported in and around Bandar Abbas during Thursday night’s U.S. strikes, and highways connecting the port city to nearby provinces were declared closed, according to Iran’s state broadcaster IRIB.

Iranian state media and a U.S. defense official said the U.S. has been striking targets throughout the country and not just along the coast. The official said the targets include weapons and surveillance systems that Iran has used to attack commercial shipping. Among the targets are small boats, coastal-radar sites, air-defense systems and missile- and drone-storage facilities.

The U.S. has repeatedly hit Chabahar, Iran’s only deep water, oceanic port. Defense Secretary Pete Hegseth posted a picture of the collapse of a maritime-communications tower in Chabahar, which sits more than 350 miles east of the strait near the Pakistan border.

Iranian authorities have confirmed the facility was struck and insisted it was used for civilian purposes, such as search-and-rescue operations for fishermen at sea. Chris Long, a former British naval officer in the Persian Gulf, said the tower could also have served as an observation and intelligence post.

“We are likewise winning big in Iran, and you will see the fruits of that labor very, very shortly,” Trump said in a prime-time address Thursday night.

Throughout the war, Tehran’s strategy has been to out-escalate and outlast the U.S. It now says it is broadening the geographical reach of its attacks in response to the intensification of U.S. attacks.

In recent days it has expanded beyond routine attacks on U.S. bases in Bahrain and Kuwait.

Overnight, Iran fired a ballistic missile at a U.S. base in Saudi Arabia after avoiding hitting the kingdom, a more sensitive target, in exchanges of fire under the interim deal to stop the fighting. U.S. troops were injured by attacks on Jordan but returned to duty, a senior U.S. official said.

On Friday, Kuwait said Iranian strikes damaged a power and desalination plant, a provocative escalation particularly at the height of summer that forced it to activate emergency plans. The country’s military said 32 drone attacks had been intercepted since early Thursday.

Iran also has begun attacking Qatar and Oman—two countries involved in efforts to find a diplomatic solution—and stepped up its attacks on shipping.

On Thursday, officials in Iraq said drones had targeted a tanker and a containership at its southern ports, as well as Iraq’s northern Kurdish region. The area’s biggest natural-gas field was shut amid credible threats it would be attacked. Iraqi Kurdish authorities said more drones were intercepted above the regional capital of Erbil on Friday.

No one has claimed the strikes in Iraq, but Iran and its local allies have frequently launched drones there.

Another vessel was attacked by an unknown projectile near the Strait of Hormuz on Friday, according to the U.K. Maritime Trade Operations, which is affiliated with the Royal Navy.

Ship tracker Kpler said traffic through Hormuz has dropped to a three-week low. Half were Iranian ships, and most of the traffic went through the Iranian route.

The next crash: why this time might not be different

Stock markets are not only ignoring the obvious threats, but seem imbued with extreme optimism

In the autumn of 1929, Irving Fisher, one of the greatest American economists, stated: “Stock prices have reached what looks like a permanently high plateau.” This turned out to be one of the most incorrect forecasts ever made: in short order, US and global stock markets were hit by the Great Crash, which was followed by the Great Depression.

Some scholars argue that Fisher was analytically right: markets were indeed valuing the US capital stock correctly in 1929. But his “permanently” proved unambiguously mistaken. Maybe markets were in some sense “right” before the crash and wrong after it. But who cared? For investors and the hundreds of millions of people across the world whose lives were upended by the disaster, the gods of the stock market had failed for a generation.

Why might this story be relevant today? The answer is that the valuation of US stocks is even higher today than in September 1929. In only one other period since 1881 has there been a higher valuation of the US market than today’s. That was in 1999-2000, immediately before the bursting of the “dotcom” bubble. (See charts.)

We know what followed that peak: a fall in valuations that was as steep, though not as deep, as the one after 1929. The first global financial crisis after the disaster of the 1930s followed a little over six years after the bursting of the 1999 bubble, in 2007-09. This was not just a coincidence: the easy monetary — and relaxed regulatory — policies adopted after the stock market bubble burst contributed to the financial crisis.

How can one best judge the valuation of stocks? The best-known metric is the cyclically adjusted price/earnings ratio developed by Robert Shiller, the Nobel laureate and Yale economist. Cape is defined as today’s value of the stock market divided by the average of the previous 10 years of earnings per share, both in real terms. Shiller also provides a “total return Cape”, which adjusts for changes over time in the payout policies of companies. There is a difference between the two calculations, but it is not large.

The intuition underlying this measure is simple: if people are paying more than has been historically normal for the underlying earnings of companies, the stock market can be judged to be relatively highly valued, and vice versa. Data for the US market, measured in the S&P 500, goes back to 1881. Over that period, the mean Cape has been 17.8, which indicates a healthy average real return of 5.6 per cent. There have also been three huge peaks: September 1929, at 32.6, December 1999, at 44.2 and, crucially, July 2026, at 41.4.

The Cape is a simple and effective measure of value. But Shiller also offers a more sophisticated one: “the excess Cape yield”. This measures the difference between the inverted Cape ratio — or cyclically adjusted aggregate earnings per share — and inflation-adjusted yields on Treasury bonds.

When stocks are expensive, the excess yield is low. When stocks are cheap, the excess yield is high. Crucially, when the excess yield is low, the subsequent 10-year excess returns on stocks have normally been poor. The excess yield is a mere 1.4 per cent in July 2026, far below its long-run average of 4.7 per cent. With stocks this expensive, the chances of healthy future returns must, again, be relatively low.

How does today’s US situation compare with those of comparable markets elsewhere? One way of addressing this question is to examine the ratio of the total value of the stock market to GDP, also known as the “Buffett indicator” (after Warren Buffett). At over 200 per cent in the US in early 2026, this was extraordinarily high by US historical standards and more than double UK levels.

High valuations of US shares add to the size and relative dynamism of the overall US economy and so of its corporate sector to make its stock market far and away the world’s most important: in June 2026, it accounted for 55 per cent of the global value of stock markets (at current prices). The US market is a titan.

Historically, when US valuations have come close to this sort of level, a crash has followed. Stocks almost always fall much more quickly than they rise. Will things be different this time?

We do know that the Cape is far from a perfect predictor of an imminent crash: if it were, it would not be; well-informed investors would then not allow markets to reach extreme positions in the first place and so crashes would be far less likely.

This time, boosters always say, is different. One justification for such optimism is the scale of the AI-related boom. Its impact includes expectations of huge profits for the “hyperscalers” (Amazon, Google, Meta, Microsoft and SpaceX) and also the suppliers of microprocessors (notably Nvidia) and memory chips. Although not yet floated, OpenAI and Anthropic are also expected to be highly rewarding, even at eye-popping valuations.

As the most recent annual economic report of the Bank for International Settlements notes, these buoyant expectations are fuelling an enormous surge in US investment, which itself raises economic optimism.

The view that artificial intelligence is a transformative technology is quite reasonable. But the history of investment surges underpinned by profound innovations does not show that the latter guarantee huge profits.

Over-investment, destructive competition, waves of bankruptcies and then painful consolidation are standard features of such episodes, from the railway booms of the 19th century to the internet boom of the 1990s. This is the classic capitalist story of booms and busts.

This matters very much for the future of today’s markets. As Chris Watling, chief executive at Longview Economics, noted in June, “a handful of AI-linked stocks accounts for roughly 40 per cent of the S&P 500’s market capitalisation, according to Bank of America data”. Thus, the current extraordinary valuations of the market depend on the continuation of the AI boom.

Given the scale of what is happening, the latter, in turn, depends on the materialisation of one (or both) of two hoped-for outcomes — faster productivity growth and/or a big shift in income from labour to capital. Neither is guaranteed. But if the former happened, real interest rates would rise, lowering the present value of the higher future earnings. If the latter happened, it would tend to create political and social upheaval — hardly the ideal environment for peaceful enjoyment of enhanced earnings.

Without such huge transformations in economic growth and the distribution of income, today’s Cape suggests prospective real returns of a mere 2.4 per cent, less than half the historic average. At some point people will realise this, and the market will crash.

One argument against such a pessimistic conclusion is that the US market has been more expensive on average since, say, 1960 than it was in the previous 80 years, perhaps because both economic management and access to index funds have improved. Thus, since 1960, Cape has averaged 21.7. This is indeed above the average of 17.8 since 1880. But that is still just about half of what it is today.

Current valuations look to be a huge stretch. Moreover, as Joachim Klement, a strategist at Panmure Liberum, wrote in the FT, not just the valuations but even the earnings themselves look to be in a bubble.

Nobody knows for certain what might trigger the corrections. But we can see plenty of opportunities for destabilising shocks in which a stock market plunge is just one part of a bigger story.

First, we have definitively lost the stabilising and benevolent US hegemon of old. Under today’s irrational and unpredictable management, anything is possible. The on-again-off-again war on Iran is the perfect example. The unpredictable trade war is another. Uncertainty has costs. Moreover, for the first time since its emergence as a superpower in the early 20th century, the US has, in China, a peer competitor.

Second, the ratio of public debt to GDP in the advanced economies is back to where it was at the end of the second world war, even though there has been no war, but a financial crisis, a pandemic and fiscal profligacy, notably that of Donald Trump, instead. According to the IMF, the US now has general government fiscal deficits of over 7 per cent of GDP. The public debt of emerging economies, though lower than that of advanced economies, is also at an all-time high.

Private debt is also worrying. Data from the Institute of International Finance shows private gross indebtedness is close to where it was on the eve of the global financial crisis. Worse, there is procyclical financial deregulation, which is precisely what tends to occur some decades after the crisis that justified the earlier tightening. This inevitably exacerbates the perils of periods of “irrational exuberance” such as today’s.

Third, high and rising debt generates financial fragility. The BIS report focuses rightly on the interaction of the rising government debt with the increasing role of hedge funds in funding that debt. The strategy of the latter depends on leverage. That increases the risks of a panic in which trades unwind at high speed. We have already seen such disruptions early in the pandemic and again in the UK’s “Truss shock” of September 2022.

Yet this is far from the only form of financial fragility. Another is the lack of transparency created by the growing role of non-bank financial intermediation, especially in the US. Another, again, is the rapid pace of financial innovation, notably the rising role of weakly regulated stablecoins. Would these be as reliable as money needs to be in a crisis? If not, flight could heap panic on panic.

Fourth, the underpinnings of dynamic market economies — the rule of law, support for science, orderly government — are under attack, notably in the US. This links with the disarray of global economic governance, notably the trading system on which our economies continue to depend.

Finally, as Manoj Pradhan and Charles Goodhart argue in The Unanchored Central Banker, the combination of de-globalisation with ageing will, in the longer term, generate still higher interest rates and rising fiscal pressures. As a result, they argue that central banks will lose their ability to “anchor” inflationary expectations.

My best guess on what might trigger the correction? Fiscal pressures, higher long-term interest rates, forced monetisation, inflation, financial shocks and panics. But war and accelerated de-globalisation are also possibilities.

The markets, above all US markets, are not only ignoring all such threats, but also embracing a highly optimistic view of the prospects even of what is going well. Stocks are, as a result, extremely expensive.

So, what are investors to do? This depends, as always, on both their time horizons and capacity for bearing losses. If the former are long and the latter are large, they can stay fully invested. Those without the luxury of time or robust financial security need to hedge. Options are a possibility; cash (and not just dollars) and precious metals are others. In today’s world, remember the downside risks.

Findings raise questions over the quality of the FCA’s information

New regulatory data released this week on short sellers’ bets against UK-listed companies contained several apparent errors, raising questions over the quality of the information being provided to the market under the new disclosure regime.

The Financial Conduct Authority published the data after new rules ended the practice of publicly naming hedge funds and investors holding significant short positions — bets that a company’s share price will fall. Instead, the regulator will now disclose the total short positions in a company on an aggregate basis.

But the new information published by the FCA contained several apparent errors, with positions subsequently removed or changed without a record, according to an analysis by data provider Breakout Point that was reviewed and confirmed by the FT.

The data also included years-old positions which are highly unlikely to still exist.

The findings raise questions over the quality of the FCA’s data, which is based on private submissions by investors about their short positions and is an important source of information for traders and regulators.

“The information that flows into the FCA is essential for market oversight and to identify misconduct,” said Chris Brennan, partner at law firm Dentons. “Market users have a reasonable expectation that what they see published is correct.”

Short positions against FTSE 250 IT infrastructure company Softcat were disclosed in an FCA report on Monday but were not included in an updated report on Tuesday. However, the change was not noted in a section setting out historic positions that have since been closed.

Tuesday’s report also listed different dates and sizes for short positions in four companies, including student accommodation provider Unite, without noting that these details differed from the information published on Monday.

One of the adjustments was made because a short seller’s position had been duplicated in the data, said a person familiar with the matter.

“Early-days issues are perhaps forgivable and things are already improving, but invisible corrections in an official market record should not become a habit,” said Ivan Cosovic, founder of Breakout Point.

The regulator’s reports on Monday and Tuesday contained other errors, with one of the documents being wrongly dated. They also appeared to omit some short positions which had been disclosed as active in the final report under the old disclosure regime, including shorts held by Saba Capital and Lombard Odier Asset Management.

Those positions had been disclosed as active last Friday but a person familiar with the matter said these positions had not been carried over in the latest disclosures as they were historic.

However, the regulator did publish details of some short bets that are more than five years old and are therefore unlikely to still be active or to be the same size as when they were previously disclosed.

For example, a short position in miner Critical Mineral Resources disclosed in April 2021 appeared to have been included on an anonymised basis in the FCA’s new report despite the company suffering an 87 per cent fall in its share price since the bet was first made public. Short sellers normally close their positions to secure a profit after a company’s share price drops sharply.

It is not known whether the potential errors were introduced by the FCA or stemmed from inaccuracies in data submitted to it by investors.

The FCA said it had considered the examples of potential errors raised by the Breakout Point analysis and concluded “there is no need for any revisions [to the data] at this point”.

It added: “The FCA monitors reported positions and engages with position holders where necessary to verify information and maintain the accuracy of published disclosures. We are also monitoring how the regime is operating and will consider whether changes are necessary.”

A person close to the FCA noted that its reports rely on the timeliness and accuracy of the information provided to it, and that it approaches firms to understand whether older reported positions remain valid.

The FCA has long published daily reports on hedge funds’ and asset managers’ short positions in UK companies.

Under the new rules, the regulator reports the total short interest in each company but it no longer discloses the size of each investor’s individual position or their identities.

The FCA will make a public disclosure when aggregate short interests exceed 0.2 per cent of a company’s share capital. It previously only published details of short positions of more than 0.5 per cent.

The spot rate for a 40-foot container from Shanghai to Los Angeles rose to $6,482 last week, the highest since 2024.

This marks the 10th consecutive weekly increase.

Spot rates for this route have nearly TRIPLED since the Iran War began in February.

Still, the Port of Los Angeles processed more than 530,500 loaded inbound containers in June, up +13% YoY, the highest volume in any June on record.

The surge has been driven by importers rushing goods into the US ahead of the expiration of temporary tariffs on July 24th and expected new import taxes, while the Iran War continues to disrupt shipping patterns.

Supply chain stress is surging again.

In the age of intelligence, how should firms protect their core IP?

Nobel Prize winning economist Kenneth Arrow famously described a paradox in the market for information. “Its value for the purchaser is not known until he has the information, but then he has in effect acquired it without cost.” In Arrow’s “Information Paradox,” the seller risks giving away knowledge in order to sell it.

AI creates the reverse problem. In the AI age, the buyer risks giving away knowledge, just in order to use what they bought.

You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!

Over time, the information asymmetry becomes increasingly skewed. The seller learns more and more about you as you use what you purchased, while you learn very little about what the seller is learning in return.

That is what I think of as the Reverse Information Paradox.

Patents solve one aspect of Arrow’s paradox. They let an inventor disclose an idea without simply giving it away. The Reverse Information Paradox needs its own equivalent.

This requires more than data protection. Models learn from "exhaust," the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong. Every correction is distilled into institutional know-how. It's the kind of knowledge a competitor could never buy, and the kind that leaks almost imperceptibly: trace by trace, correction by correction, eval by eval.

In consuming intelligence, you are creating intelligence. And what you create should belong to you. This is your particular intelligence, in Hayek's sense: the knowledge of time, place, and circumstance that no one else can hold. It knows what you think, what you value, and how you measure success.

While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and to reserve the right to learn from customer usage and interaction data. If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself. Therefore, it's imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop.

As Alex Karp put it: "What the technical customers want is control over their compute, their models, their data stack, and their alpha. They want to know they own the means of production, and it's not being transferred to someone else." The current regime does precisely the transfer Karp and companies fear.

That is why enterprises need a real trust boundary for their human capital and token capital to compound. It is where an organization’s data, traces, evals, adapted weights, and memory accumulate and improve together. And it is a hard boundary across which nothing crosses, not even the intelligence exhaust, without consent. Enterprises will demand the rights to use model outputs to fine tune and/or train their own models. I think of this as every firm’s right to align models to their enterprise accountability obligations.

In the cloud era, enterprises accumulated data. In the AI era, they accumulate learning. The trust boundary must evolve accordingly, from protecting information to protecting the mechanisms through which organizations learn, adapt, and compound intelligence. There are a few things every enterprise must do to ensure this:

Control: Create your private evals, because evals define what “good” looks like inside the organization. Also, retain ownership of your organization’s memory, traces, feedbacks, decisions, and institutional context, and ability to use outputs of models from your own tasks and queries.

Capability: Build your own proprietary learning environments within the tenant boundary to train or tune models, where models learn against real workflows without exposing the company’s knowledge.

Choice: Ensure the orchestration layer is decoupled from any single model. Ask yourself: If any one model you are using is taken away, do you still have the ability to operate and optimize for your evals using other models? Does your company “veteran” capability remain with you even if a given “generalist” model is taken away?

Cost: By decoupling the orchestration layer, you are also able to bring together context, models, and tasks in the most efficient and cost-effective way without sacrificing quality.

Compound: Bring these four together and you create your own continuous learning loop (i.e. hill climbing machine) that will allow your AI investments to compound the value of your firm.

In other words, a company should be able to use a model without giving up the knowledge that makes it unique. That is the reverse information paradox we need to confront.

Frost Investment Advisors isn’t a household name like fellow bond fund managers Pimco or BlackRock. But perhaps it should be.

The firm’s $4.2 billion Frost Total Return Bond fund has completely dominated its Morningstar intermediate core-plus bond fund category over the past three, five, 10, and 15 years, beating over 95% of its peers in each period. In a weak time for bonds overall, the fund has delivered a 6% three-year annualized return and 3.1% over five years. By comparison, the average fund in the group has produced a measly 0.3% annualized return over the past five years, and the $398 billion Vanguard Total Bond Market Index exchange-traded fund has notched an even worse minus 0.1%.

The key to the Frost fund’s success has been lead manager Jeffery Elswick’s astute application of a flexible strategy. “If our view is that the bond market is going to deliver negative returns over the next 12 months, we give ourselves large latitude to minimize that downside in whatever way we can,” he says.

Elswick’s 12-month outlook on interest rates and the economy affects how much duration risk—a measure of bond interest-rate sensitivity—he will take, as well as the kinds of bonds he will own. Although the fund can hold any U.S. bond, he limits its exposure to high-yield bonds with low credit qualities to no more than 25% of the portfolio, since junk bonds increase a fund’s correlation with the stock market. As of March 31, the fund was only 4.9% invested in high-yield bonds, as Elswick doesn’t find their yields justify the additional credit risks.

The 2022 bond downturn and the 2023 recovery show the advantages of such flexibility. At the start of 2022, the U.S. Treasury market was poor from a risk/reward perspective because of low rates, Elswick says. He shortened the duration of his portfolio to a range between 2.5 to three years while that of the typical fund in his category was five to six years. Bond prices move inversely to interest rates, and the longer a bond’s duration, the more sensitive it is to rate moves. As inflation and rates spiked in 2022, the average core-plus bond fund fell 13.3%. Frost fell only 5.5%.

“I describe 2022 to our clients as our best year in 20 years and our worst year in 20 years,” because the fund isn’t supposed to lose money, Elswick says. Yet in 2023 the fund also outperformed its peers and benchmark, rising 8.4% because of astute individual security selection and the manager having increased the duration to four years.

Currently, the fund’s duration is 5.4 years, which, “versus the prior 10 years, is on the very high end and much closer to our benchmark duration,” Elswick says, referring to the Bloomberg U.S. Aggregate Bond Index. “We’re of the view that the U.S. economy is slowly normalizing and that money-market rates are not going to change appreciably over the course of the next year.” He expects the yield on the 10-year Treasury note will largely be range-bound between 4.4% and 4.6%.

Although the Iran war is a continuing concern vis a vis inflation, which drives interest rates, oil and other commodities “are actually lower than where they were a couple of months ago,” he says. “Also, U.S. tariffs on trade peaked around October of last year, and that is a tailwind for lower inflation.” For this reason, Elswick expects the Federal Reserve either not to change interest rates in the next year or to make one small cut by year-end.

In such a range-bound environment, Elswick wants to collect the coupon yield on high-quality bonds. But instead of Treasuries, he’s favoring higher-yielding securitized debt, primarily mortgage-backed bonds guaranteed by government agencies such as Ginnie Mae. A stable-rate environment is good for mortgage bonds as declining rates lead to more homeowners prepaying their mortgages and refinancing them with newer ones at lower rates—good for homeowners but less so for investors. Meanwhile, higher rates hurt all fixed-rate bonds.

Currently, 53% of the fund is in securitized debt, versus 34% for the average fund in its category and only 21% for the Vanguard index fund, according to Morningstar. The highest-quality agency debt yields about 1.2 percentage points above comparable Treasuries.

Elswick will also hunt for value in lower-quality corporate debt—recently 17% of his portfolio. He holds beleaguered aircraft-maker Boeing’s bonds, for example, which mature in 2064 and have a coupon rate of 7.1%. But Elswick purchased the issue at a significant discount to its face value, so it yields even more.

“We’re always searching for ideas that a lot of folks just don’t want to touch,” he says. “Boeing is the perfect example of that. We started to add to Boeing when they were having problems with their new Dreamliner [aircraft]. The market really beat the name up more than what we thought was justified.”

He also recently purchased bonds from software company Oracle, which has suffered from leveraging up its balance sheet to build data centers. S&P Global gives Oracle bonds a BBB- rating —investment-grade, but only one notch above junk bonds. Yet the bonds have been trading at a discount, with higher yields equivalent to BB-rated credits.

“This is still one of the premier companies in its industry,” he says. “The management team has said pretty firmly that they want to continue to keep the company at investment-grade-rated levels.”

The fund generally doesn’t invest in derivatives, such as swaps or index futures, which larger competitors like Pimco do for liquidity purposes. Elswick doesn’t like the leverage and counterparty risks derivatives add. “If we were running a trillion dollars like a Pimco, we almost certainly would have to do something different,” he notes.

Instead, Elswick employs a team of 10 analysts and traders to scour the market for individual bonds. For this, the fund charges a 0.71% expense ratio, which is higher than the Vanguard ETF’s 0.03%. If you have a Charles Schwab account, you can buy the institutional share class, FIJEX, which has a more attractive 0.46% fee, for a $1 minimum. Elswick also says a low-fee ETF share class of the fund may be forthcoming.

Maybe that will attract more investors’ attention.