What If the AI Bears Are Right About Circular Financing?

The artificial intelligence infrastructure boom has become one of the largest capital deployment cycles we have seen in technology.

Billions of dollars are flowing into GPUs, data centers, power generation, networking, cooling, and the supporting infrastructure required to train and operate increasingly sophisticated AI models. The demand is real, the technological progress is remarkable, and there are legitimate reasons to believe that artificial intelligence will reshape large portions of the global economy.

But there is another side of this investment cycle that deserves more attention.

My associate Rich Washburn recently wrote an excellent article examining in depth what he describes as the AI industry's emerging “company store”—an increasingly interconnected system in which technology companies invest in AI companies, those companies purchase or contract for enormous amounts of computing capacity, infrastructure providers raise capital to build that capacity, and much of that capital eventually finds its way back to the companies supplying the GPUs and technology that made the investment cycle possible in the first place.

It raises a question I have been thinking about: What if the AI bears are right about some of the financing, even if they are wrong about AI?

I am not referring to skeptics who believe artificial intelligence is another passing technology trend. The evidence increasingly suggests otherwise. I am more interested in the short sellers, traders, credit investors, and contrarian institutional investors who accept the importance of AI but question whether the extraordinary amount of capital being deployed around it can ultimately generate adequate economic returns.

That is a much more interesting argument.

Follow the Money

At a simplified level, the AI economy increasingly operates as an interconnected capital ecosystem. Large technology companies invest in AI developers. Those developers require enormous amounts of computing capacity. Cloud companies and specialized infrastructure operators raise debt and equity to construct data centers and acquire GPUs.

Those purchases generate revenue for semiconductor and infrastructure suppliers, while greater access to compute allows AI companies to expand their models, attract customers, raise additional capital, and commit to still more infrastructure.

There is nothing inherently troubling about this structure. Strategic investment, vendor financing, equipment financing, and long-term customer commitments have existed across industries for decades. In many cases, these arrangements are precisely what allow new technologies and infrastructure markets to scale.

The issue for investors is not whether the transactions are legitimate. The more important question is: whether the structure makes underlying demand more difficult to measure.

When a supplier is also an investor, when a customer is simultaneously backed by companies that benefit from its infrastructure spending, or when an infrastructure provider can raise substantial amounts of capital based on contracts with a relatively concentrated group of AI customers, traditional measures of demand require more scrutiny. Revenue remains revenue and a GPU shipment remains a GPU shipment, but investors still need to understand where the economic demand ultimately originates.

This is where I believe the contrarian argument deserves to be taken seriously.

Being Right About the Technology Isn't Enough

Technology history provides plenty of examples of investors correctly identifying a transformational technology while still losing enormous amounts of capital financing its development.

The telecommunications boom surrounding the early internet is perhaps the most relevant comparison. Investors correctly anticipated extraordinary growth in internet traffic and telecommunications demand. Billions of dollars were deployed into fiber networks, networking equipment, data centers, and communications infrastructure. Equipment manufacturers sometimes financed customers purchasing their products, capital was readily available, and companies raced to build capacity before their competitors could.

The fundamental technology thesis was overwhelmingly correct. The internet changed the global economy more profoundly than even many of its early advocates anticipated.

The capital allocation thesis was considerably less reliable.

Too much capacity was built in some markets, leverage became excessive, pricing assumptions proved optimistic, and numerous companies failed despite owning infrastructure that eventually became extremely valuable. The lesson wasn't that the internet had been overhyped.

The lesson was that a revolutionary technology does not automatically make every investment supporting that technology a good investment.

Artificial intelligence could produce a similar distinction. AI may become one of the most important technologies of our lifetime while portions of the infrastructure supporting it are simultaneously overbuilt, overleveraged, or financed at returns that ultimately prove inadequate. Those two outcomes can coexist.

The Useful Life of a GPU Changes the Equation

AI infrastructure also introduces an interesting capital question because the assets inside a modern data center have dramatically different economic lives. The building itself may remain productive for decades. Electrical infrastructure, transformers, switchgear, cooling systems, and power-generation assets are generally designed around long useful lives. Much of that infrastructure can also retain value regardless of which generation of computing equipment occupies the facility. GPUs operate on a very different technology cycle.

The leading AI accelerator today may remain economically useful for years, but its relative value can change quickly as newer generations deliver better performance, greater energy efficiency, and improved economics. That creates a potentially important mismatch between long-duration infrastructure financing and rapidly evolving computing hardware.

For lenders and equity investors, the relevant question is therefore not simply whether a data center has customers today. It is whether the economics remain attractive through hardware refresh cycles, changes in compute pricing, shifts in utilization, and potentially significant changes in power requirements.

This is why utilization, counterparty quality, customer concentration, power costs, hardware residual values, and financing terms matter as much as the headline number of GPUs installed or megawatts announced. The physical infrastructure may have a thirty-year life. The economic assumptions supporting it may not.

Time to the Pay the Piper

For all the complexity surrounding AI financing, I think the ultimate test is surprisingly simple. Eventually, businesses outside the AI capital ecosystem need to generate enough economic value from artificial intelligence to support the infrastructure being built for them.

A pharmaceutical company needs to accelerate drug discovery. A manufacturer needs to increase production efficiency. A financial institution needs to reduce operating costs or improve decision-making. A logistics company needs to move goods more efficiently. A software company needs its employees to produce materially more output. Consumers need to find AI applications valuable enough to pay for them directly or indirectly.

Those are the dollars that ultimately matter because they represent economic value entering the AI ecosystem rather than capital circulating within it.

If artificial intelligence creates enough productivity across the broader economy, today's extraordinary infrastructure spending could ultimately prove entirely rational. We may look back and conclude that the industry was racing to build the foundational infrastructure for a technological transformation comparable to electrification, telecommunications, or the internet.

But if downstream economic value develops more slowly than the infrastructure being financed upstream, the consequences could be very different. Compute pricing could fall, utilization assumptions could disappoint, weaker operators could struggle to refinance equipment, and investors could discover that demand for AI was real while their assumptions about the economics of providing that AI infrastructure were too aggressive.

That would not mean artificial intelligence failed. It would mean capital got ahead of economics.

Where Eliakim Capital Fits

At Eliakim Capital, we remain constructive on the long-term need for AI infrastructure, however we believe the current environment requires more discipline rather than less.

Compute cannot be evaluated independently from power, power cannot be separated from capital, and capital ultimately cannot be separated from the economics of the customer consuming the compute.

That interconnectedness creates substantial opportunities, particularly for operators that can secure the right hardware, reliable power, appropriate financing, and credible customers. It also makes underwriting more important as increasingly large amounts of capital pursue the same secular theme.

The largest data center is not necessarily the best investment. The company with the most GPUs is not necessarily the strongest operator. A multibillion-dollar backlog is only as valuable as the counterparties behind it and the economics under which those commitments were made. And the ability to raise billions of dollars does not, by itself, establish that those billions can be deployed at attractive returns.

I don't think investors need to choose between being an AI bull or an AI bear. That framing is probably too simplistic for what is happening.

The bulls may ultimately be right about the technology, while the bears may still be right about portions of the financing. The more important task is understanding where one ends and the other begins.

For investors and operators allocating capital into AI infrastructure today, that distinction may prove far more important than predicting the next generation of GPU.

____________________________________________________________________________

Todd Colpron is the Managing Partner of Eliakim Capital, a private investment and strategic advisory firm that invests its own capital and co-invests with a family office.

Important Disclaimer

This publication is provided solely for general educational and informational purposes and reflects the author’s views and analysis as of the date of publication. Nothing contained herein constitutes, or should be construed as, investment, financial, legal, accounting, or tax advice; a recommendation or endorsement of any security, issuer, transaction, financing arrangement, or investment strategy; or personalized advice based on any reader’s particular circumstances.

This publication does not constitute an offer to sell, a solicitation of an offer to buy, or an invitation to enter into any transaction involving securities or other financial instruments. Any offer or sale of securities may be made only through applicable offering documents, by appropriately authorized parties, and in accordance with applicable federal and state securities laws.

Eliakim Capital is a private investment and strategic advisory firm that invests its own capital and co-invests with a family office. Eliakim Capital is not a registered broker-dealer nor registered investment adviser and does not represent itself as providing brokerage, securities execution, investment management, or other services requiring registration or licensure under applicable securities laws. Eliakim Capital does not accept or maintain custody of investor funds or securities, maintain customer brokerage accounts, or execute securities transactions on behalf of readers or the general public.

Unless separately and expressly disclosed in a written engagement, Eliakim Capital is not acting as an investment adviser, broker, dealer, placement agent, fiduciary, attorney, accountant, or other licensed professional to any reader. No investment advisory, brokerage, fiduciary, attorney-client, accountant-client, or other professional relationship is created by accessing or reading this publication. Nothing herein should be interpreted as representing that Eliakim Capital or any of its principals holds any securities license, registration, or professional credential not expressly identified by Eliakim Capital.

Information contained in this publication may be derived from public filings, company disclosures, press releases, industry sources, third-party publications, interviews, publicly available information, or other materials believed to be reliable. Neither Eliakim Capital nor the author makes any representation or warranty, express or implied, regarding the accuracy, completeness, timeliness, or reliability of such information, and information presented herein may not have been independently verified. Opinions, estimates, interpretations, and conclusions are made as of the publication date and are subject to change without notice.

Certain statements contained herein may be forward-looking and may include expectations, estimates, projections, assumptions, or opinions regarding anticipated events, technologies, business prospects, financial performance, valuations, market opportunities, financing conditions, industry developments, or future results.

Forward-looking statements are inherently uncertain and subject to known and unknown risks, assumptions, and other factors that could cause actual events or results to differ materially from those expressed or implied. Neither Eliakim Capital nor the author undertakes any obligation to update or revise such statements as a result of new information, future events, or otherwise.

References to companies, issuers, securities, technologies, transactions, markets, valuations, financing arrangements, or industry participants are provided solely for informational and analytical purposes and should not be interpreted as an endorsement, recommendation, or indication that any particular investment or transaction is appropriate for any person.

Investments in private companies, emerging and growth-stage businesses, small-cap and microcap companies, technology ventures, infrastructure projects, and other alternative or less-liquid investments may involve substantial risk. Such investments may be speculative, volatile, illiquid, subject to dilution, dependent upon additional financing, and based on limited publicly available information. Investors may lose some or all of their invested capital.

Readers should conduct their own independent investigation and consult appropriately licensed financial, legal, tax, accounting, and other professional advisers before making any investment, financial, legal, tax, or business decision.

Past performance is not indicative of future results. No representation, warranty, projection, or assurance is made regarding the future performance or outcome of any investment, company, security, transaction, strategy, technology, project, or market discussed herein.