Matthew Sigel, VanEck’s head of digital assets research and the manager of the VanEck Onchain Economy ETF (NODE), said he does not see the current AI infrastructure boom as a bubble. In his view, the more immediate drag on crypto is not macro pressure but institutional disappointment with the adoption trajectory of major layer-1 networks.
Sigel made the remarks on The Rollup podcast episode “AI Super Cycle,” hosted by Rob (Robbie Klages) and Andy and aired on Aug. 10, 2026. In the discussion, he spoke about the market’s shift away from rewarding capital expenditure, the rerating of Bitcoin miners as AI data-center plays, the comparison between AI buildout and the 19th-century railroad bubble, and why VanEck has reduced exposure to mainstream L1 tokens.
The program also included a disclosure that Sigel serves as VanEck’s head of digital assets research and as the manager of NODE, meaning the views expressed may align with positions held by VanEck funds.
From rewarding capex to punishing it
Sigel said the first five months of the year were marked by an unusually concentrated market structure: the companies spending the most on capital expenditure were the ones whose stocks performed best. That changed on June 1.
Since then, he said, the relationship has flipped. Leaving aside the rebound that followed the forced unwind at Situational Awareness, the highest-spending companies were hit the hardest once market leadership turned.
He said Bitcoin and crypto tokens are effectively being grouped with software assets in the current market regime. Investors have been watching the relative performance of semiconductors and software, and Bitcoin, in his framing, is software — open-source software, specifically. He added that AI tools such as Claude and Codex are already having a real effect on many open-source projects, while those projects do not enjoy the same top-down upgrade cycle that Web2 companies can impose. Users cannot simply be forced to upgrade. That broader weakness across software, he said, has weighed heavily on Bitcoin and crypto tokens.
Sigel also pointed to the psychological grip of the four-year halving cycle. He said he would rather see greater dispersion across assets than a market driven by one dominant factor. In that kind of environment, he said, investors have a better chance of generating alpha from specific stocks and specific assets instead of making one broad thematic bet.
As for his own positioning, Sigel said he still gets AI infrastructure exposure through Bitcoin miners. He said he is constructive on a fourth-quarter bottom for Bitcoin and expects the market to rebalance, but he would rather wait for bullish catalysts with volume behind them than trade every move in a choppy market.
Why NODE leaned hard into miners pivoting to AI
When asked about holdings such as MARA, Riot, APLD, and WULF, Sigel said NODE has been live for 15 months and has outperformed Bitcoin by nearly 100 percentage points in that span. The biggest contributor, he said, was recognizing early that the valuation attached to each megawatt controlled by Bitcoin miners was far too low relative to the multiples assigned to the small number of existing data-center REITs at the time.
He described the old mining model as a capital-intensive, low-margin business built on repeated dilution. Miners issued equity to buy ASICs and had to move faster than competitors just to stay in the race, while revenue was effectively cut in half every four years. He called that a melting-ice-cube business.
What changed, he said, is that financing conditions around AI data-center construction improved sharply. This was not limited to hyperscalers themselves; lease financing around those facilities also became much cheaper. That opened the debt market to miners, allowing them to raise capital without constantly issuing stock and diluting shareholders.
Sigel said the economics have improved with each new lease signed, and lower rates add more value on top of that. Because of this, the style rotation that began in June hurt the most leveraged investors first. VanEck, he said, did not use leverage and tried to stay light on highly levered companies.
He gave one example. When Situational Awareness, a fund with large overlap to VanEck’s holdings, was liquidated by its prime broker, VanEck bought into the weakness instead of stepping back. On the opening last Thursday, he said, the team made the largest single-day trade in the fund’s history, moving nearly 10% out of low-volatility, low-beta positions and doubling down on the miners it liked most.
Sigel said VanEck has not seen any fundamental deterioration in returns on the hyperscaler AI spend. If anything, he said, earnings calls from companies such as Amazon indicate returns have been better than originally expected. He cited older GPUs that had previously been rented out at $2 per hour and were later renewed at sharply higher pricing once the original contracts expired. For him, that is why fundamentals come first, and those fundamentals are still improving.
At the lows, he said, some of VanEck’s highest-conviction names were valued only on existing leases, with no credit given to data-center terminal value, platform value, or potential future leases. Even under an assumption that the 10-year yield rose another 100 basis points, he said the work had already been done, which left a large margin of safety at those levels.
No systemic threat to Bitcoin from miners moving into AI
On the question of whether shifting computing capacity from Bitcoin mining to AI could create a systemic problem for the Bitcoin network, Sigel said no.
He argued that a lower hash rate would actually make the remaining miners more profitable. He also said companies VanEck added to, including Bit Deer and MARA, still retain meaningful strategic flexibility: they can continue mining, or they can repurpose facilities toward AI workloads.
There may be a future Bitcoin price that reopens the debate over whether some operators should switch back to mining, he said, but VanEck is not calling for AI-converted facilities to revert now. His estimate was that CleanSpark would need Bitcoin to reach $360,000 before it made economic sense to tear up freshly signed AI leases and go back to pure mining. That optionality, in his view, is valuable on its own.
Why he says this is not a railroad-style bubble
Sigel also addressed a comparison he has heard repeatedly: that AI infrastructure resembles the U.S. railroad bubble of the 19th century, a transformative capex cycle that still destroyed early capital.
His answer was that the comparison breaks down on both scale and financing structure.
On scale, he said the United States spent roughly 3% of GDP on railroads for nearly 20 consecutive years. AI, by contrast, is only now hitting that 3% of GDP level for one year, even though it has been about five years since GPT emerged. He added that current AI-company valuations are not pricing a 20-year buildout. In fact, he said, most analysts think the cycle peaks in 2030. That mismatch is the first disconnect in the bubble argument.
On financing, Sigel said the railroad bubble was government-led. He pointed to the 1862 legislation under which Congress granted hundreds of millions of acres of federal land to railroad companies, while the actual title to that land would only transfer after the network was completed. The U.S. Treasury also issued construction bonds, and those bonds were subordinate to private capital. Large railroad companies then sold debt overseas as if it were a safe asset, even though repayment leaned on land sales tied to property the companies did not yet own. In his telling, that is what made it a bubble.
AI factories, he said, are financed on very different terms. Railroads needed a complete route before their utility fully emerged. An AI facility can start generating value as soon as it is connected to the grid, linked to fiber, and equipped with chips. It can train models and run inference right away without waiting for an entire global network to be built.
He also stressed the difference in backlog. Nobody in 1870 was buying a train ticket for 1885, and there was no forward freight market to support those projects. A great deal of the speculation revolved around future land sales. Today’s data-center market looks different, he said. The four biggest cloud providers have more than $2 trillion in contracted backlog, with Microsoft and Oracle accounting for roughly half of that total. Many of those contracts include customer prepayments, customer-supplied GPUs, and terms longer than five years. In other words, the compute these facilities are expected to produce is already backed by real purchase commitments, and financing is being raised against those contracts rather than government subsidy.
For that reason, Sigel said the present AI buildout is more durable. It rests on long-term private-sector agreements and years of backlog that lenders and investors can underwrite.
VanEck cut major L1 exposure and turned toward enterprise chains
Sigel’s comments on crypto were just as direct. He said VanEck reduced its mainstream L1 exposure after the U.S. election, when many tokens doubled but actual adoption did not accelerate in step, no breakout application emerged, and no application drew in global capital in a way that changed the category.
What VanEck saw instead, he said, was the rise of enterprise chains. He named Circle, Stripe, and Robinhood as examples and described these networks as semi-permissioned systems that let public companies tailor the user experience and capture part of the economic upside.
He acknowledged that this may not fit the open-source ethos or crypto purist thinking, but said large-scale users care about predictable fee flows. On that point, he brought up ETH before Solana, saying Ethereum serves as a clear example of how volatile transaction costs can become and why large institutional participants often want more stable operating costs.
Solana, he said, has less fee volatility, but until recently one still could not buy USDC on Solana in New York. He then pointed to fresh reporting that Wells Fargo was studying its own tokenized deposit chain. To Sigel, this reflects a broader reality: banks and other regulated entities do not want to place meaningful capital directly on open public chains. Even when they do participate, they may choose to support three to five different chains at once, which weakens the winner-take-all case for any single L1.
That is why, he said, VanEck remains very underweight these tokens. In his view, market share is leaking from open L1 networks toward enterprise chains, and the winner-take-all dynamic that once underpinned many token theses is becoming less convincing.
CLARITY Act could trigger a relief rally, but he remains cautious
Sigel said one policy development could still change sentiment: the CLARITY Act.
If the bill passes, even though he said the probability of passage has fallen to its lowest point of the year, some tokens could see a large relief rally. The reason is disclosure. A formal disclosure regime, he said, would let the market identify the true beneficial owners behind token positions and reveal how much is held by labs and foundations.
That would make it harder for key opinion leaders to promote tokens without disclosing their own positions. More broadly, Sigel said the lack of a disclosure framework is one of the reasons many institutional investors have chosen to ignore the sector altogether. Until that changes, he said, he intends to stay cautious.
Lower inflation is sensible, but details will matter
Toward the end of the episode, the hosts asked about proposals from Ethereum, Solana, and Near to reduce validator inflation. If approved, the inflationary supply of some major L1s could move close to zero.
Sigel said VanEck studied this question a few quarters ago by comparing average inflation rates across major L1s with user growth and fee revenue. His conclusion was that the sector does have an inflation problem.
He linked that issue back to the broader software trade. Semiconductors have substantially outperformed software this year, he said, and that has pushed many software companies to rethink how aggressively they issue new shares. Some have already reduced dilution. For L1s, he said, revisiting token inflation after six years is a sensible move.
He did, however, caution that second-order effects could be meaningful, especially for companies whose business models depend on staking income. He singled out Bitmine, saying the company frequently promotes its staking yield and that this piece of the model could come under pressure if inflation falls. On balance, he said, lower inflation is the right direction, but implementation details will determine whether it works.
The discussion was sourced from The Rollup podcast and republished in summarized form by Deep Tide TechFlow. The original episode title was “VanEck Research Head: Why This Isn't The AI Bubble Everyone Fears (Here's Why),” and it aired on Aug. 10, 2026. WuBlockchain’s repost note said the material was shared for information purposes only and did not constitute investment advice or represent WuBlockchain’s own position.