TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Podcast Notes | VanEck Head of Digital Assets Research: Current AI Infrastructure Market Is Not a Bubble; Crypto Market Sluggishness Stems from Institutional Disappointment with L1s

Podcast Notes | VanEck Head of Digital Assets Research: Current AI Infrastructure Market Is Not a Bubble; Crypto Market Sluggishness Stems from Institutional Disappointment with L1s

2026.08.11
Share
He correctly predicted the transformation of crypto mining companies in advance; the ETFs he manages outperformed BTC by nearly 100 points; his judgment is worth watching.

Organized & Compiled: TechFlow

Guest: Matthew Sigel, Head of Digital Assets Research at VanEck, Fund Manager of VanEck Onchain Economy ETF (NODE)

Hosts: Rob (Robbie Klages) & Andy, The Rollup "AI Super Cycle"

Podcast Source: The Rollup

Original Title: VanEck Research Head: Why This Isn't The AI Bubble Everyone Fears (Here's Why)

Air Date: August 10, 2026

Disclosure: Matthew Sigel is the Head of Digital Assets Research at VanEck and Fund Manager of NODE. The views expressed in this episode may align with the positions held by VanEck funds.

Key Takeaways

The VanEck Onchain Economy ETF (NODE) managed by Matthew Sigel has outperformed Bitcoin by nearly 100 percentage points over the past 15 months, primarily due to an early bet on the transformation of Bitcoin mining stocks into AI data centers. He believes that for the first 5 months of this year, the market was highly concentrated: whoever spent the most on capital expenditure saw their stock price rise the most; but after June, this logic reversed, software assets (including Bitcoin and crypto tokens) were sold off together, and capital expenditure itself became the target of punishment.

Sigel's stance can be summarized in two judgments: First, AI infrastructure is not a bubble, it differs from the 19th-century railway bubble because this round of financing comes from long-term leases in the private sector, rather than government land grants and speculative bonds;

Second, the real pressure on the crypto market is not macroeconomic, but institutional disappointment with mainstream L1s. Since the election, VanEck has reduced holdings in mainstream L1s such as Solana and ETH, shifting focus to enterprise chains like Circle, Stripe, and Robinhood. He believes that if the CLARITY Act passes and establishes an information disclosure system, related tokens may see a huge relief rally, but until then, he will remain cautious.

Highlights

On the Shift in Market Structure

  • "For the first 5 months of this year, the companies that spent the most performed the best; but starting in June, the companies that spent the most were punished the hardest."
  • "Bitcoin is software, coincidentally open-source software. When software stocks are under pressure overall, it is difficult for Bitcoin and crypto tokens to stand alone."
  • "I hope to see more differentiation in the market, rather than a single factor deciding everything. This way we have the opportunity to find alpha from individual assets."

On the AI Transformation of Bitcoin Mining Stocks

  • "ASICs are not the miner's most valuable assets; electricity and land are."
  • "The early miner's business model was constantly issuing shares to buy mining machines, with revenue halving every four years; this is a melting ice cube. Now they can finance through the debt market without diluting shareholders."
  • "For CleanSpark to return to a pure mining mode, Bitcoin would have to rise to $360,000, because only then would it be worth tearing up the AI leases already signed."

On the Comparison Between AI Infrastructure and the Railway Bubble

  • "Back then, the U.S. poured 3% of its GDP into railways for 20 consecutive years, while AI has only reached this level for one year so far."
  • "The railway bubble was government-led: Congress passed a bill in 1862, granting land first, with transfer of ownership only after the network was built; construction bonds issued by the Treasury were even subordinated to private capital."
  • "AI factories can train models and perform inference immediately once connected to the grid, fiber optics are installed, and chips are in place. Railways, however, were only truly useful after the East Coast was connected to the West Coast."
  • "The four major cloud providers hold over $2 trillion in contract backlog, with many contracts involving customer prepayments and customers bringing their own GPUs, with terms exceeding five years. In 1870, no one would buy a train ticket for a railway opening in 1885."

On L1s and Enterprise Chains

  • "After the election, many tokens doubled, but no truly breakout applications emerged, nor did any application attract global capital."
  • "What is winning now is enterprise chains: Circle has one, Stripe has one, Robinhood has one, and even Wells Fargo is building its own custom chain."
  • "Banks and other regulated institutions do not want to put real money on open-source chains; even if they support them, they must support three to five chains simultaneously, which dilutes the winner-take-all attribute of any single L1."
  • "If the CLARITY Act passes, bringing a real information disclosure system, some tokens will see a huge relief rally. Until then, we remain cautious."

Full Text

Chapter 1: Market Shifts from "Chasing Capital Expenditure" to "Punishing Capital Expenditure"

Rob: Tell us about the market state you've seen recently. There was a period before where regardless of positive news like DTCC or the CLARITY Act, digital assets didn't react positively. Now Saylor is selling coins, CLARITY looks like it won't pass, and the market is neither rising nor falling, completely indifferent. At the same time, AI and U.S. stocks seem to be sucking away all liquidity. Is this what you see?

Matthew Sigel: Yes. For the first 5 months of this year, the market was very concentrated, and the companies that spent the most saw their stock prices rise the most. But starting June 1, this relationship reversed. Putting aside the recent rebound from the bottom of the Situational Awareness liquidation, during that period, the companies with the highest capital expenditure fell the hardest.

Bitcoin and crypto tokens are actually categorized as "software." Many people watch the relative strength of semiconductors and software, and Bitcoin is software, specifically open-source software. AI capabilities like Claude and Codex are having a substantial impact on many open-source software projects, but their upgrade cycles are not top-down and changeable at will like Web2 companies. You cannot force users to upgrade. This situation where software overall underperforms largely weighs on Bitcoin and crypto tokens.

Coupled with the psychology of the four-year halving cycle, the market is very awed by this model. I hope the market will show more differentiation and more dispersion of returns in the future, giving us the opportunity to find alpha from individual stocks and individual assets, rather than relying solely on betting on a certain factor. I don't think it must be that AI rises and crypto falls, or vice versa; reality will be more subtle than this.

As for my own positions, I remain exposed to the AI infrastructure theme through Bitcoin miners. I am optimistic about Bitcoin bottoming in Q4 and also expect the market to rebalance. However, I tend to wait for positive news accompanied by volume before acting; repeatedly trading short-term in volatile markets is not my style.

Chapter 2: Why NODE Bet Heavily on the Miner AI Transformation

Rob: So what are your specific positions now? I see some tickers, like MARA, Riot, APLD, WULF, which are essentially Bitcoin mining stocks transforming into AI data centers. They have already built infrastructure, have electricity and energy capacity, often in remote areas. These happen to be the raw materials needed for AI data centers, and existing hardware can even be switched in use. How long do you think this trend can last? Also, if so much computing power flows from the Bitcoin network to AI, will it bring systemic problems to Bitcoin?

Matthew Sigel: The NODE ETF I manage has been established for 15 months and has outperformed Bitcoin by nearly 100 percentage points. The largest source of return was the early discovery: every megawatt of electricity controlled by Bitcoin miners was severely undervalued relative to the valuation multiples of the few data center REITs at the time.

The early miner's business model was constantly diluting shareholders, issuing stock to buy ASICs, just to run faster than competitors. But revenue halves every four years; this is a melting ice cube, capital intensive with thin profits. Now we can see that not only hyperscalers, but even the construction leases around these data centers, financing costs have dropped significantly. These companies can finance through the debt market without continuing to issue stock to dilute shareholders. Moreover, every new lease signed has economic conditions almost better than the previous one. Coupled with declining interest rates, this creates significant value.

So the market style shift since June hurt highly leveraged people the most, and we didn't use leverage and tried to underweight highly leveraged companies. When we saw funds like Situational Awareness, which highly overlapped with our holdings, being liquidated by prime brokers, we chose to add positions against the market. At the opening last Thursday, we made the largest single-day trade since the fund's inception: withdrawing nearly 10% of low-volatility, low-beta positions and doubling down on the few miners we favor most.

We have not seen a fundamental deterioration in the capital return on this investment by hyperscalers. In fact, reading earnings conference calls from companies like Amazon, these returns are even better than early expectations. Old GPUs rented out for $2 per hour before, now after contracts expire, customers want to renew at significantly higher prices.

So we look at fundamentals first, and fundamentals are still improving.

When it fell to low points, the stock prices of some of our most favored companies only reflected the value of existing leases, completely without pricing the terminal value of data centers, platform value, and potential new leases. Even assuming the 10-year yield rises another 100 basis points, this work has already been done. So at the low points, we saw a large margin of safety.

Regarding the Bitcoin network, I do not think there are systemic problems. On the contrary, after the hash rate declines, the miners who remain earn more. Companies like Bit Deer and MARA that we added positions to still have strong optionality: continue mining, or convert facilities to AI. Some future Bitcoin price might make people re-discuss whether to switch back to mining, but we are not calling for facilities already converted to AI to switch back. We estimate CleanSpark would need Bitcoin to rise to $360,000 before it would be worth tearing up the newly signed AI leases. So this optionality itself is very valuable.

Chapter 3: This Is Not a Bubble, This Is the Counterexample to 19th-Century Railways

Rob: You mentioned historical analogies. You wrote an article comparing AI infrastructure to 19th-century railway construction. Can you quickly explain why this comparison holds?

Matthew Sigel: Many people use the railway bubble to refute me, saying this is also a capital expenditure cycle that changes the economy, but early capital was largely destroyed. My response is: you can compare scale, and you can compare financing structure.

In terms of scale, the U.S. poured 3% of its GDP into railways for nearly 20 consecutive years back then. On the AI side, counting from the appearance of GPT to now is the fifth year, and only this year has it reached the level of 3% of GDP. Based on current valuations of AI companies, the market is not at all pricing in "this round of construction will last twenty years." Most analysts believe it will peak in 2030. This is the first disconnect.

Second is the financing structure. The railway bubble was actually government-led. In 1862, Congress passed the Railway Act, granting hundreds of millions of acres of federal land to railway companies, but land ownership was only transferred after the entire network was built. Moreover, the Treasury issued construction bonds, and these bonds were even subordinated to private capital. The largest railway companies went overseas to sell bonds, marketing them as safe assets, and these bonds relied on land grants, but at the time the companies didn't even have land ownership. This was the bubble.

Compare this to AI factories. Railways needed to be connected to a network to generate real utility; value was limited before the East and West Coasts were connected; AI factories can train models and perform inference immediately once connected to the power grid, fiber optics are in place, and chips are racked. There is no need to wait for the global network to be completed synchronously.

The last important difference is contract backlog. In 1870, no one would buy a train ticket for a railway opening in 1885; there was no forward freight market at the time, and everything was built on speculation of land sales. But for current data centers, just the four major cloud providers have over $2 trillion in signed contract backlog, with Microsoft and Oracle accounting for about half combined. Many contracts include customer prepayments, customers bringing their own GPUs, and terms exceeding five years. The computing power produced by these factories is supported by real purchase orders, and financing is also based on these contracts, not government subsidies.

Rob: So the conclusion is, this round is more sustainable.

Matthew Sigel: Yes, more sustainable. It is supported by long-term contracts in the private sector and has years of backlog as collateral.

Chapter 4: Mainstream L1s Are Losing to Enterprise Chains

Rob: Someone in the livestream mentioned, was it 2021 or 2023, you said Solana needed to pause L1 development. If Solana wants you to look favorably on it again, what does it need to do? Also, we see many L1s reducing inflation, reducing incentives to validators; Ethereum just released a new proposal this morning, and Solana and Near are too. How important are these inflation adjustments to L1 market conditions?

Matthew Sigel: After the election, when many tokens doubled but actual adoption did not accelerate and no truly breakout applications emerged, we reduced the company's L1 exposure. What we saw afterwards was the rise of enterprise chains. Circle has one, Stripe has one, Robinhood has one; these are all semi-permissioned chains, allowing listed companies to customize user experience and extract partial economic benefits.

I understand this does not align with the open-source spirit and crypto fundamentalism, but those who want to use these networks on a large scale need predictable fee flows. You asked about Solana, but I want to mention ETH first, because ETH is an example of transaction fees fluctuating too much, and many large institutional participants want more stable costs. So enterprise chains have captured a lot of market share, and this is also the focus of our investment attention.

Solana's fee volatility is not that large, but until recently you couldn't buy USDC on Solana in New York; this morning we again saw Wells Fargo researching its own tokenized deposit chain. Banks and regulated entities do not want to put real money on open-source chains; even if participating, they must support three to five L1s simultaneously, which dilutes the winner-take-all attribute of a single chain. So market share is being lost, and the winner-take-all characteristic is also weakening, which makes us very underweight on these tokens.

If the CLARITY Act can pass, although the probability has dropped to the lowest of the year, I think some tokens will see a huge relief rally. It will establish an information disclosure system, letting us know who the real beneficial owners are, and there won't be KOLs shilling without disclosing holdings. We can also see how many tokens labs and foundations hold respectively. It is precisely the lack of a disclosure system that makes many institutional investors choose to ignore this field. I hope to see this change, and only then can we seriously look favorably on many L1s again.

Chapter 5: Reducing Inflation Is Good, But Details Determine Success or Failure

Rob: ETH, Solana, and Near have all proposed reducing validator inflation; if passed, the inflationary supply of these L1s might approach zero. How important do you think these measures are to the market?

Matthew Sigel: A few quarters ago we conducted a study comparing the average inflation rate of mainstream L1s with user growth and fee revenue. There is indeed an inflation problem in this field. At the beginning of this conversation, we compared software companies and crypto tokens; this year semiconductors significantly outperformed software, forcing many software companies to reflect on the speed of issuing new shares, and the dilution level of some companies has already decreased significantly.

So I think it is appropriate for L1s to explore reducing inflation. But details will determine success or failure, and we must also look at the second-order effects on companies relying on staking revenue, for example, Bitmine often promotes its staking returns, and this part might collapse. Generally speaking, it is wise for L1s to rethink inflation six years after their birth.