Roundhill just launched the Neocloud ETF, and rumors immediately emerged that 'The Big Short' is shorting its top holding: who’s right?

The AI computing power market witnessed a rather dramatic scene.

On August 6, Roundhill officially launched its Neocloud-themed ETF— $Roundhill Neocloud ETF (NCLD.US)$ , aiming to bundle $NEBIUS (NBIS.US)$ 、 $CoreWeave (CRWV.US)$ along with a group of AI data center companies that have pivoted from cryptocurrency mining, into a tradable 'AI computing infrastructure asset.'

Almost simultaneously, news emerged that famed short-seller Michael Burry had established a short position in NEBIUS at $211.77 per share. According to media reports, due to the put options on NBIS having implied volatility exceeding 100%, Burry opted to short the stock directly rather than buying put options. Following the news, $NEBIUS (NBIS.US)$ the stock dropped roughly 13% in a single day, $CoreWeave (CRWV.US)$ 、 $IREN Ltd (IREN.US)$ 、 $Cipher Digital (CIFR.US)$ and $TeraWulf (WULF.US)$ and other Neocloud-related stocks also weakened collectively.

More noteworthy than the rumor itself is why this piece of news could so quickly trigger a sell-off across the entire sector.

The answer lies in the fact that it struck right at Neocloud’s most sensitive current issue: the market doesn’t doubt there’s still demand for AI computing power—the real debate is whether these companies can convert massive orders into sufficiently high returns for shareholders.

What exactly does NCLD invest in?

$Roundhill Neocloud ETF (NCLD.US)$ It is an actively managed ETF that began trading on Nasdaq on August 6, with an expense ratio of 0.65% and initial assets of approximately $1 million. Roundhill defines Neocloud as companies whose primary revenue, orders, or capital expenditures are tied to GPU-as-a-Service, AI data centers, high-density colocation, power, and high-speed networking infrastructure.

From its first-day holdings, NCLD exhibits extremely high concentration: $NEBIUS (NBIS.US)$ and $CoreWeave (CRWV.US)$ The top two holdings together account for 58.13%, and the top five holdings exceed 77% of the portfolio. Therefore, although NCLD is nominally an ETF, it essentially functions more like a 'NEBIUS + CoreWeave dual-leader portfolio,' supplemented by a group of highly leveraged names—former Bitcoin miners transitioning into AI data centers. It offers a high-purity exposure to Neocloud rather than genuine risk diversification.

Symbols

Roundhill’s bullish thesis rests primarily on three assumptions:First,Demand for AI training and inference continues to outpace compute supply;Second,Neocloud companies can lock in revenue through multi-year compute contracts;Third,Large customer contracts can serve as a financing foundation, enabling companies to build additional compute capacity without excessively diluting shareholders.

Morgan Stanley forecasts that global data center construction spending could reach approximately $2.9 trillion by 2028. Neocloud aims to become the 'compute tollbooth' connecting NVIDIA GPUs, data center capital, and AI customers.

This is a classic 'build first, harvest later' logic.The question lies in how much capital must be invested upfront and whether revenues can materialize in time before capital costs spiral out of control.

The controversy surrounding NEBIUS: growth is extremely rapid, but spending is even faster

$NEBIUS (NBIS.US)$ First-quarter revenue reached $399 million, up 684% year-over-year; adjusted EBITDA turned positive at $129.5 million, reflecting rapidly materializing demand for computing power and operating leverage.

On the other hand, the company’s capital expenditures on property, plant, equipment, and intangible assets in the quarter amounted to $24.73 billion—more than six times its revenue for the same period. Depreciation and amortization totaled $2.12 billion, or approximately 53% of revenue, resulting in an operating loss of $128 million.

As of the end of Q1, NEBIUS held approximately $9.3 billion in cash, but its non-current debt also stood at $8.432 billion. The company’s 2025 annual report further disclosed that it had signed—but not yet commenced—data center leases with undiscounted future payments totaling approximately $9.76 billion.

This does not necessarily indicate an imminent financial crisis for NEBIUS, but rather highlights that the Neocloud business model carries both high operating leverage and high financial leverage.

If demand for computing power remains tight, customer contracts are fulfilled as scheduled, and GPU utilization stays elevated, revenue growth could quickly outpace depreciation and financing costs, turning prior investments into stronger profitability.

However, if computing power prices decline, data centers experience delivery delays, or major customers reduce demand, the pre-committed costs for GPUs, power, and leases could shift from competitive advantages to cash flow pressures.

Therefore, what the market is truly debating isn’t whether NEBIUS has orders, but three more critical questions: when those orders will convert into revenue, whether that revenue can cover the cost of capital, and ultimately how much of the returns will actually accrue to common shareholders.

Bulls’ counterargument: Can orders translate into financing capacity?

NEBIUS is currently not relying solely on equity issuance to fund its expansion.

The company’s deferred revenue for the first quarter rose to approximately $4.78 billion, reflecting that some large clients have already supported capacity build-out through advance payments. In July, NEBIUS also completed its first secured debt financing of $775 million at a rate of SOFR plus 2.5%, backed by GPU assets and cash flows from investment-grade customer contracts.

The company stated that this financing, together with cash flows from customer contracts, is sufficient to cover more than 100% of the capital expenditures related to the associated GPU deployments.

This is the most critical component of the bullish thesis.

If long-term contracts from major clients like Microsoft and Meta can be recognized by financial institutions as high-quality collateral, NEBIUS could establish a closed funding loop—'sign contracts → secure project financing → deploy GPUs → recognize revenue'—reducing reliance on equity issuance.

In that case, high capital expenditures might simply represent upfront investments preceding a revenue surge, rather than a permanent cash drain.

However, a single $775 million financing round is still insufficient to fully validate that this model can support future infrastructure needs amounting to tens of billions of dollars. The market will now need to verify whether NEBIUS can replicate this type of financing repeatedly at similar costs, rather than executing just one-off projects.

Why was NCLD launched at this time?

The launch of NCLD indicates that Neocloud is gradually evolving from a handful of highly volatile stocks into a standalone AI investment theme. However, the ETF’s listing itself does not mean Wall Street has fully endorsed this business model.

Its initial assets under management are only around $1 million, and the top two holdings account for over 58% of the portfolio—risk is not meaningfully diversified merely by virtue of the ETF structure. NCLD is more of a high-beta trading vehicle betting on AI compute expansion, rather than a low-volatility, broad-based allocation product.

What truly needs to be watched next is:Whether Nebius and CoreWeave’s committed computing capacity can be deployed on schedule; whether orders and deferred revenue can be successfully converted into actual revenue; whether the pace of capital expenditure growth can gradually fall below revenue growth; and whether GPU depreciation, financing costs, and equity dilution can decline.

Nebius will report its second-quarter results on August 12 (U.S. Eastern Time), which could serve as the first key catalyst for the market to reassess the bullish and bearish arguments around Neocloud.

The more reasonable conclusion at this point is that AI computing demand may remain strong, but having the right direction does not mean any valuation is justified. Roundhill is betting that the compute supply gap will persist, while bears question whether capital costs will arrive sooner than profits.

Ultimately, what determines Neocloud’s value is not the size of its orders, but whether the company can convert those orders into sustainable free cash flow before the financing window tightens.

617K Views