When the Shoe Company Becomes an AI Company

Allbirds’ pivot to AI is more than a quirky headline. It is a warning shot about narrative-driven valuations, fragile capital structures, and the growing credit risk inside the AI infrastructure boom.

KEY TAKEAWAYS

• Allbirds became a cult sneaker brand with a multibillion-dollar valuation before pivoting toward an AI identity, signaling how quickly narrative can override core business fundamentals.

• AI is a legitimate, once-in-a-generation technological shift with massive long-term economic impact, but speculative financing structures are building around it.

• The AI infrastructure buildout is expected to require trillions in capital across data centers, power, chips, and supporting systems, pushing heavy demand into fixed income and increasingly into private credit markets.

• Large, established companies have the balance sheets, cash flow, and diversification to support aggressive AI spending, while outsourced compute providers are far more dependent on leverage and narrow demand assumptions.

• The CLEC telecom bubble offers the more relevant comparison: real demand and real infrastructure paired with excessive leverage and poor capital discipline that ultimately led to widespread destruction.

MY HOT TAKES

• The market is starting to price the word “AI” more aggressively than the actual economics behind the business. That is usually when discipline leaves the building and future regret walks in.

• Private credit is becoming one of the most important hidden variables in the AI buildout story. If the marginal dollars funding this boom come from a stressed credit market, the risk is bigger than most equity investors appreciate.

• Complexity in bond indentures is not a sign of sophistication. It is often a sign that the underlying economics need makeup to get financed.

• The biggest AI winners may still be the same dominant platforms with cash, scale, and durable revenue streams. The casualties are more likely to come from the over leveraged imitators trying to rent their way into relevance.

• A technology can change the world and still destroy a large number of investors along the way. Being right on the trend does not protect anyone from being wrong on valuation, timing, or capital structure.

• You can quote me: “The technology was not wrong. The capital structure was wrong.”

Dot AI. Ok, right off the bat, I have to admit, I wasn’t ready to pen this missive. It was–of course–floating around in the back of my head as a one-of-these-days-when-the-timing’s-right piece. 🌀But, alas, I am compelled. I spent the better part of my younger years pursuing all manner of sports–particular endurance sports. I thought–at the time–that it was healthy for my body and, at the time, it was! I liked nothing better than a good late night run with a family member or friend. A 2-mile swim in a freezing cold pool at 4 AM will certainly wake you up and get you ready to tackle the day. And nothing beats the deep thinking that can be done on a 60-mile bike ride–but for the hills, it truly is relaxing. All of this on top of team and racket sports have left my over-fifty ankles, knees, shoulders, and back in a–let’s call it–slightly-used condition. That condition is not compatible with the dress code of my position, which demands that I not wear the athleisure wear worn by most millennials, gen-Zs, and even some of my gen-X cohorts. In other words–no sneakers. Then came Allbirds!

Allbirds was founded in 2015 by former professional soccer player and a renewable resources expert with a singular mission: to build a better sneaker using natural materials instead of plastics and petroleum-based products. The company became a cult favorite among the Silicon Valley tech crowd, known for its minimalist wool and eucalyptus fiber shoes that managed to be both environmentally conscious and genuinely comfortable. At its peak, Allbirds carried a Wall Street valuation north of $4 billion. For me, it was a windfall! For some reason, those wooly-sneakers became acceptable in the C-suite. First on dressdown Fridays and ultimately on any day suits were not required. My ankles, knees, and hips are grateful to the founders. Unfortunately like all trends, they can fade fast–especially when Silicon Valley has something to do with their success. 😉And to be fair, the sneakers were not exactly the coolest looking bit of garb. That hasn’t stopped me, the now late-fifties pragmatist.

Yesterday, I was minding my own business attempting to navigate one of the most challenging, but exciting, economic, market, and geopolitical regimes in GENERATIONS. And then I saw it. Allbirds–the savior of my soles (I had to get that Dad-joke in 🤣)--announced that it is transitioning into a–I can’t make this up–AI company that will be called NewBird AI. My first response? “Wait what?” And then within literal seconds: “ruh roh!”

You see, I have lived through this before. In fact, my long-time, faithful followers know that I was on the front lines of the dot-com bubble working on mergers and acquisitions. You could say that I know a thing or two about asset bubble valuations and what could go wrong if you drink too much Kool-aid. Now, I don’t want to in any way pass judgement on NewBird AI’s move without the necessary diligence, but the signal is undeniable. And to be clear, it didn’t necessarily come out of the blue for me. I have been noticing an increasing amount of investment “opportunities” passing over my desk from companies attempting to raise money in the public capital markets attempting to rebrand themselves as AI infrastructure companies. They all come with super-rich valuations based on what they refer to as no-brainer, high-probability-for-success opportunity. The first part–no-brainer–I give them. The latter–high probability–maybe not so much.

Now layer this on top of what I have been writing about an awful lot recently. Of course, I am referring to private credit, more broadly the credit markets. Let me start with this: AI is awesome and a once-in-many-generations sea change that will likely propel US economic growth into a new multi-decade boom phase. Similar, but bigger than the PC revolution–more akin to the industrial revolution and steam engines. Got it? Great. But realizing this great potential is going to take A LOT OF INVESTMENT. Thankfully, some of the companies that were forged in the PC revolution and internet revolution have amassed knowledge and capital war chests that can get us on that path. You know which companies I am talking about. They are in the process of spending massive–MASSIVE amounts of capital to build out data centers, power infrastructure, silicon fabs, etc. They have cash and that’s good. But cash alone won’t nearly fit the bill. The most logical place to raise capital to be used for capital projects is the fixed income market. And companies are definitely taking advantage of those markets and their mostly high credit ratings. So far, so good. This is not a bubble and the companies borrowing the money are healthy and they can afford the debt. But something changed a bit last summer when a growing number of investors began to question the explosive growth in CAPEX. I am not sure if that was the catalyst, but I noticed that the bond indentures raising capital for infrastructure became increasingly complex. Then came a few crypto mining companies transitioning into AI data center outsourcing. AI companies quickly realized that they could offload CAPEX from their models by outsourcing to these compute-renters and push the CAPEX burden on them. Why own the cow when you can rent it? These outsourcers are building massive data centers and compute infrastructures to satisfy contracts to companies like Meta and Alphabet. How are they paying for all that? Debt, of course. This reminds me–and I all over the public record on this–of the CLEC bubble which really caused the Dotcom bubble to burst.

Let’s start with some history. The CLECs–Competitive Local Exchange Carriers–were the darlings of the late 1990s telecom buildout. Deregulation had thrown open the doors, capital was cheap and plentiful, and every entrepreneur with a fiber spool and a PowerPoint deck was promising to rewire America. The narrative was irresistible: bandwidth demand was doubling every hundred days, the internet was going to consume every morsel of available capacity, and whoever laid the most pipe would own the future. Sound familiar? The CLECs borrowed massively, built furiously, and then watched in horror as supply overwhelmed demand virtually overnight. Bandwidth prices collapsed by some 90%! Companies that had been market darlings filed for bankruptcy in waves. The industry owed roughly a trillion dollars, much of which was simply written off. Bond investors recovered about twenty cents on the dollar. The infrastructure was real. The demand was real. The debt, however, was catastrophic, and it brought down companies that had no business being in the infrastructure game in the first place, alongside some that probably did.

Now multiply that by three. Morgan Stanley and Moody's estimate that at least $3 trillion in capital spending will be required to build out the data centers and related infrastructure needed to support the AI boom, and JPMorgan puts the number north of $5 trillion once power generation is included. That, my friends, is not a rounding error. That is the largest coordinated infrastructure buildout in the history of capital markets, and it has to be financed. The most logical place to raise that kind of capital is the fixed income markets, and companies are absolutely doing that with investment grade issuers borrowing against strong balance sheets, high credit ratings, and real cash flows. So far, so good. But here is where it starts to get uncomfortable. A growing share of that capital is not coming from the blue-chip bond market. It is coming from…wait for it…wait for it…private credit–the very same $1.8 trillion market where retail investors have been recently rushing for the exits, where Moody's just revised its outlook on non-traded BDCs to negative, and where redemption gates are going up at KKR, Blue Owl, and many others across the space. The same market, in other words, that was already showing signs of stress before anyone stapled the word "AI" to a convertible note and called it an infrastructure company. When the marginal financing for a $3 - $5 trillion buildout is coming from a credit market that is itself under pressure, that is not a footnote. That is the story underneath the story.

Here is the part that the headlines always miss when they invoke the Pets.com example everybody loves to cite. Pets.com was silly. A sock puppet mascot selling fifty-pound bags of dog food and shipping them at a loss was never going to work, and anyone with a pencil and a napkin could have figured that out in about four minutes. The CLEC story is more instructive precisely because the underlying thesis was correct. The internet did transform everything. Bandwidth demand did explode. The fiber that was laid in a frenzy of speculative overbuild did eventually get used–actually, every bit of it and then some. YouTube exists 🎉 because bandwidth became essentially free after the crash. Streaming exists. Cloud computing exists. The infrastructure winners of the next generation were built on the wreckage of the infrastructure losers of that one. The technology was not wrong. The capital structure was wrong. The valuations were wrong. And critically, the identity of the players who had any business being in the game was wrong.

That distinction is the one that matters right now, and it is the one that the market is almost completely ignoring.

Let me be precise about where I see the fault line. The Amazons, the Microsofts, the Metas, the Alphabets, the IBMs–these are not the companies keeping me up at night. These are enterprises with real, diversified, cash-generating businesses that were built over decades and that will continue to generate revenue regardless of what happens to AI infrastructure valuations next year or the year after. When Microsoft borrows in the bond market to build a data center, it is doing so from a position of extraordinary financial strength, with a balance sheet that can absorb shocks and a business model that does not live or die on AI infrastructure utilization rates. The same logic applies across that group. They are spending massively–almost incomprehensibly so–but they have earned the right to make that bet. Their debt is real debt, yes, but it is debt that the underlying business can service even in a stressed scenario. That is an entirely different animal from what I am describing.

The animal I am describing is the one that just showed up in my inbox–wearing wool. 🐑

The companies that concern me are the ones whose entire existence is predicated on a single thesis: that AI compute demand will remain insatiable, that utilization rates will stay high, that the hyperscalers will keep renewing and expanding their outsourcing contracts, and that capital markets will remain open and accommodating indefinitely. These are the compute-renters, the GPU-as-a-service shops, the data center outsourcers who took on the capital expenditure burden that the Metas and Alphabets were wise enough to offload. Their entire value proposition rests on a single arbitrage: borrow at 10% - 15%, build infrastructure that depreciates the moment the next NVIDIA chip generation ships, and pray that one or two anchor customers never decide to bring those workloads back in-house. When your entire capital structure is a leveraged bet on the perpetual goodwill of a handful of hyperscalers, you are not running a business. You are running a carry trade with a GPU farm as collateral.

And as I noted, the bond indentures began to get more complex last summer. That is not an accident. When lawyers start burying things in the covenants, it is because the underlying economics require more creative structuring to attract capital that might otherwise take a pass. I have seen this movie. The complexity is not sophistication. The complexity is camouflage (and you can quote me on that 😉).

Now layer on top of this what is happening in the equity markets, and you begin to see the full picture. When a company that sold its shoe brand for $39 million–less than ten cents on the dollar of its peak valuation–can add a $127 million of market capitalization in a single trading session simply by announcing a pivot to GPU leasing, the market is not pricing risk. It is pricing narrative. It is pricing the word "AI" the same way it once priced the word "blockchain" and before that the suffix ".com." This is not analysis. This is pattern-matching on a buzzword by investors who have watched AI-adjacent stocks go parabolic and do not want to miss the next leg. The signal is not subtle.

And here is where the credit story and the equity story converge in a way that should give every fixed income investor pause. The tight spreads we are seeing across AI infrastructure credit do not reflect the true risk profile of these borrowers. Spread compression in a hot sector is not new. It happens every cycle, in every bubble, at precisely the moment when differentiation matters most and the market is doing the least of it. When a legitimate operator like CoreWeave –a company with real revenue, real NVIDIA relationships, and a real backlog–is still carrying bonds yielding north of 11%, that is the market telling you something about the structural risk in this space. Now ask yourself what the appropriate yield is for NewBird AI, which sold its IP for $39 million three weeks ago and has never purchased a single GPU. The answer is that no serious fixed income investor should be anywhere near it at any yield, but the equity market just handed it a $150 million market cap and a platform to raise $50 million in convertible debt. That convertible paper will find buyers. It always does, right up until it doesn't.

The burst will come. I cannot tell you when, and I will not pretend otherwise. These cycles always run longer than the skeptics expect, fueled by the very real underlying demand that keeps the narrative alive and keeps the capital flowing. The internet genuinely was transformative. AI genuinely is transformative. The steam engine genuinely did change everything. None of that protected the speculators who over-leveraged themselves to build infrastructure ahead of demand, or the investors who handed them money at valuations that assumed perfection forever. History is not kind to the bagholders, even when the technology they bet on eventually wins.

What history does show is that the survivors of these cycles share a few characteristics. Strong balance sheets that can absorb a prolonged period of stress without a trip to the bankruptcy court. Management teams that have been through cycles before and know that capital discipline is not a weakness but a survival skill. Business models with enough diversification that a softening in one area does not become an existential event. Revenue streams that are real, recurring, and not wholly dependent on the continuation of a single macro thesis. These are not complicated criteria. They are, however, criteria that get thrown out the window when everyone is rushing through the same door at the same time–which is precisely what is happening now.

So where does this leave my comfortable, slightly unfashionable woolly shoes? I genuinely do not know. The Allbirds brand has been sold to a company that makes Aerosoles and Ed Hardy products, which is either a perfectly reasonable brand management outcome or a fate worse than bankruptcy depending on your fashion sensibilities. My ankles will mourn the loss either way. Thankfully, there is a well-established precedent for what to do when markets run hot, rationality takes a back seat, and the signals start flashing. You do what I have been doing since long before any of these companies existed. You keep the freezer full of ice. It has been a reliable source of comfort through more cycles than I care to count–and unlike GPU-as-a-service, the demand for it has never once disappointed. 🧊🧊

YESTERDAY’S MARKETS

Yesterday, the S&P 500 and the Nasdaq both closed at fresh all-time highs, advancing 0.80% and 1.59% respectively, while the Dow slipped by -0.15% as markets continued to price in the prospect of a US-Iran peace deal. Oil remained elevated but eased slightly, with WTI crude settling just under $91 per barrel as the Strait of Hormuz remained restricted despite ceasefire optimism. Bank earnings provided a tailwind, with Morgan Stanley surging more than 4% on strong quarterly results, while Bank of America added nearly 2%.

NEXT UP

• Initial Jobless Claims (April 11th) came in at 207k, less than last week’s 218k claims.

• Industrial Production (March) may have increased by 0.1% after climbing by 0.2% in the prior read.

• Fed speakers expected to speak today: Williams and Miran

• Important earnings today: US Bancorp, KeyCorp, PepsiCo, Citizen Financial, Bank of New York, Abbott Labs, Prologis, Netflix, and Alcoa.