Inflection Meets Liquidity

 

The Weekender is my bi-weekly take on macro shifts and emerging themes. It’s not investment advice — or even our firm’s official view. I aim simply to inform, challenge, and maybe entertain. If you’d like this in your inbox every other Saturday morning via Northern Trust, subscribe to The Weekender.

What a week.

I can't remember signals this densely packed — and we're still in August. We've had Nvidia's founder assert that "compute is revenue," perhaps the most important evolution of the AI story to date. Interest in digital payments has revived ahead of the G20 Finance Ministers meeting in Asheville, North Carolina. U.S. Treasury Secretary Scott Bessent’s recent efforts to push down long-term borrowing costs have seemingly mapped a new playbook for bond management. And Stanley Druckenmiller, one of macro's greatest traders, published an op-ed that delivered a real-time positioning lesson dressed as a warning to Bessent about respecting price discipline in bonds. There is a common thread running through much of it: liquidity.

But let's start with the most important development of the week: what Nvidia told us about AI's present, not its future.

Inflection: Compute is Revenue

For me, the most important takeaway from Nvidia this week wasn't its revenue beating expectations or its strong growth projections. It was founder and CEO Jensen Huang's assertion that "compute is revenue" — which represents a fundamental shift in how we should think about AI capital expenditure (capex) and returns. While the market worries about whether AI infrastructure capex will ever generate a return, the insiders are already saying it has.

“AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.”
– Jensen Huang

What took over a decade in the dotcom era (creating massive duration mismatches along the way) appears to have arrived for AI today. We can see returns in: the revenue growth of frontier labs like OpenAI and Anthropic; the order books and book-to-bill ratios of the hyperscalers; enterprise spend patterns, where July saw approximately a 50% month-over-month increase in AI spend per employee among the top 1% of enterprises (according to the Ramp AI Index); and forward guidance from those inside the infrastructure. My colleague Deb Koch believes cloud revenue growth could exceed capex growth next year — a shift from an industry buying graphics processing units (GPUs) in anticipation of future profits to one where inference already generates them.  

Once compute becomes a productive asset rather than speculative expense, demand becomes less cyclical, older GPUs retain value longer, capex becomes more self-funded, and AI infrastructure begins to resemble a toll road rather than a technology purchase. Capex is then simply the cost to control the future. It builds the rails into the AI economy — an economy that consensus expects will see cash-flow stall near term but then roar back to previous peaks by 2028, and then double again by the end of the decade (see this chart in this Financial Times article). As mentioned in the previous Weekender, the window to buy that at a discount closes the moment the market starts pricing the 2028–2030 free cash flow explosion, and that repricing might be happening already. See Warren Buffett's latest stock purchase for clues.

We're Still Early in the AI Cycle

So we continue to believe we are early in this AI cycle and Huang's assertion is simply the next expression, the next innings, of the cycle’s evolution. This remains true at a time when sentiment is still remarkably skeptical. Perhaps that is a product of our collective memories, where the scars of past busts continue to shape perceptions of today's boom. That's healthy. 

For me, the best sentiment signal remains that people still tell me to read 1873 by Liaquat Ahamed as a cautionary tale rather than Abundance by Ezra Klein and Derek Thompson as a roadmap. Until that changes, I suspect we are still early. After all, the crowd reading about bubbles is rarely the one inflating them.

Pay Attention to Market Anomalies

The most useful market signals rarely come from the headlines themselves, but from how markets respond to them. Especially when the response looks anomalous: when something happens that shouldn't, or when something doesn't happen that should. Pay attention, for example, when a stock goes up on bad news. It's often signaling inflection.

This week the bond market received bad news. Stanley Druckenmiller wrote an op-ed in the Wall Street Journal highlighting the unsustainable path of U.S. government deficits and warning his former acolyte, Secretary Bessent, against trying to "pick a fight with price." Druckenmiller urged policymakers to instead "let the bond market speak."

The bond market then rallied.

That's an anomaly worth paying attention to.

The Fed's Underappreciated Third Mandate

Druckenmiller is right to warn against governments trying to control bond markets. History suggests it usually ends badly. There is one notable exception: In 1942 to 1951 — a period like now where the debt-to-GDP ratio was >100% — when the Federal Reserve pursued yield-curve-control, effectively capped long-dated government bond yields and allowed nominal growth to do much of the heavy lifting, debt-to-GDP fell sharply and the economy boomed. We all know what followed: the Golden Age.

Today's environment is clearly different, but some investors argue elements of recent Treasury policy, from buybacks to bill issuance, foreign exchange intervention and liquidity support mechanisms, resemble softer forms of yield management or financial repression (also known as rocket fuel for scarce assets like gold, silver and bitcoin). Others view them simply as prudent market operations. Either way, they are worth watching.

What makes this conversation particularly interesting is that the Federal Reserve Reform Act of 1977 lists three objectives: maximum employment, stable prices, and moderate long-term interest rates. The third mandate receives remarkably little attention. Usually because the first two are enough. Could that objective become more important if the other two prove difficult to manage simultaneously? I think it could.

What took over a decade in the dotcom era (creating massive duration mismatches along the way) appears to have arrived for AI today. We can see returns in: the revenue growth of frontier labs like OpenAI and Anthropic; the order books and book-to-bill ratios of the hyperscalers; enterprise spend patterns, where July saw approximately a 50% month-over-month increase in AI spend per employee among the top 1% of enterprises (according to the Ramp AI Index); and forward guidance from those inside the infrastructure. My colleague Deb Koch believes cloud revenue growth could exceed capex growth next year — a shift from an industry buying graphics processing units (GPUs) in anticipation of future profits to one where inference already generates them.  

Once compute becomes a productive asset rather than speculative expense, demand becomes less cyclical, older GPUs retain value longer, capex becomes more self-funded, and AI infrastructure begins to resemble a toll road rather than a technology purchase. Capex is then simply the cost to control the future. It builds the rails into the AI economy — an economy that consensus expects will see cash-flow stall near term but then roar back to previous peaks by 2028, and then double again by the end of the decade (see this chart in this Financial Times article). As mentioned in the previous Weekender, the window to buy that at a discount closes the moment the market starts pricing the 2028–2030 free cash flow explosion, and that repricing might be happening already. See Warren Buffett's latest stock purchase for clues.

We're Still Early in the AI Cycle

So we continue to believe we are early in this AI cycle and Huang's assertion is simply the next expression, the next innings, of the cycle’s evolution. This remains true at a time when sentiment is still remarkably skeptical. Perhaps that is a product of our collective memories, where the scars of past busts continue to shape perceptions of today's boom. That's healthy. 

For me, the best sentiment signal remains that people still tell me to read 1873 by Liaquat Ahamed as a cautionary tale rather than Abundance by Ezra Klein and Derek Thompson as a roadmap. Until that changes, I suspect we are still early. After all, the crowd reading about bubbles is rarely the one inflating them.

Pay Attention to Market Anomalies

The most useful market signals rarely come from the headlines themselves, but from how markets respond to them. Especially when the response looks anomalous: when something happens that shouldn't, or when something doesn't happen that should. Pay attention, for example, when a stock goes up on bad news. It's often signaling inflection.

This week the bond market received bad news. Stanley Druckenmiller wrote an op-ed in the Wall Street Journal highlighting the unsustainable path of U.S. government deficits and warning his former acolyte, Secretary Bessent, against trying to "pick a fight with price." Druckenmiller urged policymakers to instead "let the bond market speak."

The bond market then rallied.

That's an anomaly worth paying attention to.

The Fed's Underappreciated Third Mandate

Druckenmiller is right to warn against governments trying to control bond markets. History suggests it usually ends badly. There is one notable exception: In 1942 to 1951 — a period like now where the debt-to-GDP ratio was >100% — when the Federal Reserve pursued yield-curve-control, effectively capped long-dated government bond yields and allowed nominal growth to do much of the heavy lifting, debt-to-GDP fell sharply and the economy boomed. We all know what followed: the Golden Age.

Today's environment is clearly different, but some investors argue elements of recent Treasury policy, from buybacks to bill issuance, foreign exchange intervention and liquidity support mechanisms, resemble softer forms of yield management or financial repression (also known as rocket fuel for scarce assets like gold, silver and bitcoin). Others view them simply as prudent market operations. Either way, they are worth watching.

What makes this conversation particularly interesting is that the Federal Reserve Reform Act of 1977 lists three objectives: maximum employment, stable prices, and moderate long-term interest rates. The third mandate receives remarkably little attention. Usually because the first two are enough. Could that objective become more important if the other two prove difficult to manage simultaneously? I think it could.

Asymmetric Information

Bessent's August 20 CNBC comments about having "asymmetric information" and his suggestion that the federal deficit may have "peaked" raise a question: Does he view the Fed’s third mandate as relevant to current conditions? I don't claim to know the answer. This would represent a fundamental shift in monetary policy and likely require reforming the Treasury-Fed Accord 1951 before implementation. But it's worth understanding the precedent, and I suspect this won't be the last we hear of it, given the reality the U.S. now pays more in interest than defense, meaning it pays more to service its past than to defend its future.

Liquidity: The Operating Framework Beneath It All

If there's one thing every global macro hedge fund manager obsesses over, it's liquidity. As Druckenmiller would say, "It's liquidity that moves markets." What I find genuinely curious in this environment is that two of the most important people in finance — Treasury Secretary Bessent and Fed Chair Kevin Warsh — are not only former global macro managers but also worked for Druckenmiller. Bessent's views are now visible in his Treasury actions. Warsh's are less obvious, put perhaps not as opaque as they seem. In 2018, Warsh co-authored an op-ed in the Wall Street Journal that began: "Around Oct. 1, global central-bank liquidity reversed; stocks began their descent from peak prices. That is no coincidence." The piece went on to question the real-economy benefits of quantitative easing (QE) (and warned that quantitative tightening would have corresponding costs.

Now, in the absence of better information, the best guide to a policymaker's future reaction function is often the framework they have already revealed. The framework Warsh revealed in his op-ed strikes me as more pragmatic and less hawkish than many assume. It’s a framework that pays close attention to productivity, liquidity and real-time data rather than mechanically tightening policy based on backward-looking inflation measures, which Warsh himself has described as "echoes of history."

Stimulus Through Stealth

Bond buybacks don't add reserves — they're not QE. But if Treasury uses the Treasury General Account for purchases, that would move money held at the Fed into markets, potentially supporting liquidity conditions. The same logic applies to Treasury bill issuance (which functions as money-good collateral in repo markets) and recent banking deregulations: Supplementary Leverage Ratio (SLR) reforms, streamlined capital requirements and similar measures allow banks to re-leverage balance sheets more efficiently. Higher balance-sheet capacity drives system liquidity higher — and with it, risk appetite (so they say…).

This is, I believe, one of the more underappreciated drivers of the current bull market. Policy may be easing through stealth. Issuing bills to buy duration creates a framework that could support a more accommodative liquidity regime over time, particularly if combined with banking deregulation, bank re-leveraging and growing stablecoin demand for Treasury bills.

And for clues about the broader liquidity backdrop, one need only glance at M2 money supply, which has recently accelerated to new highs.

A rising tide lifts all boats — something Treasury Secretary Bessent knows all too well.

Stablecoins — A Signal Worth Watching

I've written this on Thursday, before hearing from Kevin Warsh at Jackson Hole. But does anyone else find it a little odd that Jackson Hole's theme is Financial Innovation: Implications for Payments and Policy rather than growth, inflation or productivity? And that it happens to coincide with a G20 Finance Ministers meeting focused on digital assets, tokenization and cross-border payments? All the while, stablecoins are emerging as increasingly important holders of U.S. Treasury bills.

If Bitcoin's dormancy is ending and the next rally pulls more users into on-chain finance, stablecoins will likely see both higher velocity and higher issuance (in settlement). The Treasury-backed stablecoin ecosystem is essentially a distributed demand source for Treasury bills — which, given current Treasury issuance needs, is incredibly useful.

Coincidence? Perhaps. But certainly noteworthy.

The Risk: Where Inflection Meets Reality

For over a year we have argued that the biggest risk to AI may ultimately be political. Indeed, data center permitting is hitting political resistance. More than 60% of Americans now oppose new data centers in their area, up from 49% in March (according to a University of Pennsylvania survey). Prominent politicians, notably Texas Governor Greg Abbott, have reversed earlier enthusiasm for hyperscaler investment. Still, the likely landing zone appears to be converging from opposite directions on a pragmatic framework. One that allows construction but with guardrails on data handling, cost recovery on power and water (data centers pay their own utility costs), full local taxation, and community benefits packages up front. This is frustrating for hyperscalers but probably inevitable and therefore manageable. Still, it is worth monitoring.

The Data Inflection

From Warsh's 2018 op-ed: "Labor markets are a lagging indicator" and "if data dependence is the Fed's new mantra, it should actually incorporate recent data into its forthcoming policy decision." If you apply that logic today, TruFlation readings come in closer to 2%. Real-time data suggests inflation is lower than historical averages imply.

Worth noting: Prof. Raj Chetty — who pioneered big-data approaches in economics — is co-chair of the Fed's Data Task Force. The intellectual framework for "data-driven" policy is shifting toward real-time, high-frequency indicators rather than lagging surveys. That could be another signal that Warsh's pragmatism (not ideology) will shape monetary policy going forward.

Whether the next inflection comes from AI, liquidity or digital dollars, one thing seems clear: we live in fascinating times.

Have a wonderful week (and may the All Blacks reclaim their rightful spot at World No.1 after their match in Cape Town tomorrow. Come on!!!).

Gary

 

Glossary

1. Free Cash Flow: Free cash flow is cash that would be available to a company’s investors after the company has made all investments necessary to maintain the company as an ongoing enterprise.
2. Graphics Processing Unit (GPU): A specialized computer chip designed to perform many calculations simultaneously. Originally developed to process graphics and images, GPUs are now widely used to power artificial intelligence applications and other computing-intensive workloads.
3. Hyperscalers: Large technology companies that operate global cloud computing and data center networks, providing the infrastructure that powers cloud services and many artificial intelligence applications.
4. M2 Money Supply: A broad measure of money circulating in the U.S. economy. Investors often use changes in M2 as an indicator of liquidity and financial conditions because rising money supply can support economic activity and risk assets.
5. On-chain Finance: Financial transactions and activities that are recorded and executed on a blockchain network.
6. Quantitative Easing: With quantitative easing, a central bank purchases longer-term securities from the open market in order to increase the money supply, encourage lending and investment and stimulate the economy.
7. Quantitative Tightening: Quantitative tightening is a contractionary policy the Federal Reserve uses to decrease the amount of money in the economy by selling government bonds, which increases interest rates and helps control inflation.
8. Stablecoin: A stablecoin is a type of digital asset/token designed to maintain a stable value by pegging it to a reserve asset, such as the U.S. dollar.
9. Supplementary Leverage Ratio: Supplementary leverage ratio is a measurement of a bank’s Tier 1 capital relative to its total leverage. U.S. regulators set minimum required SLR ratios.
10. Tokenization: The process of representing ownership of an asset as a digital token on a blockchain, enabling ownership to be recorded, transferred, and tracked electronically.