Why Fundamentals Are Finally Winning and Where This DeFi Investor's Putting His Money | The Thesis
Summary
Kyle’s central call is that crypto is finally rewarding businesses with persistent usage, revenue and tokenholder alignment, with Hyperliquid as the proof case. Meme coins offer “no ceiling” but also “no floor”; once their differentiators blur, investors return to usage, revenue, TAM and growth. He expects a K-shaped market where roughly 90% of investments die and the strongest 10%, including Hyperliquid and Ethena, absorb the flows.
The investable shift is being reinforced by reflexivity: Hyperliquid’s success makes investors demand buybacks, better disclosure and stronger token economics from everyone else. Its roughly 99% buyback model changed expectations, while investor-relations tools and reports from Blockworks and DeFiLlama can force projects to expose weak accounting or “siphoning of money.” Kyle’s analogy is the Mag Seven: growth attracts capital, rising value strengthens the business, and the flywheel compounds.
Kyle sees selected digital-asset treasuries as distribution vehicles, not merely leveraged token wrappers. Stablecoin X—whose Nasdaq ticker USDE overlaps confusingly with Ethena’s USDe stablecoin—can give traditional investors who cannot hold ENA exposure to it, introduce Ethena’s dollars to public markets and potentially package DeFi yield. His broader pattern is that a hyped category can fall roughly 90%, after which growing fundamentals may drive a second wave to new highs.
Founder judgment ultimately reduces to whether someone can translate intellect and determination into a real business: “They have to get it, and you have to get it that they get it.” Kyle puts Ethena’s Guy Young and Hyperliquid’s Jeff Yan in that category. Nomatic said he became less bullish on Circle because he doubts Jeremy Allaire “gets it” in the same way. The hosts added Derive’s Nick Forster as evidence that persistence and responsiveness can take share from an incumbent.
Kyle’s preferred AI trade is inference demand, expressed most directly through Venice’s VVV rather than generic “AI money” narratives. Locking staked VVV creates DM, with each DM providing $1 of Venice usage per day “forever” across its models; demand for DM therefore locks VVV, while revenue burns some VVV. Nomatic’s reservation is the equity raise and three-way value-accrual question across equity, VVV and DM, though Kyle believes the market increasingly treats VVV as crypto’s clean AI proxy.
His sector filter requires product-market fit plus demand connected to an evergreen economy outside crypto. He ranks inference, tokenized stocks and DATs as S-tier, adds privacy despite admitting he owns zero Zcash, and places perp DEXs and stablecoins in A-tier. The operative lesson is timing: Mirror offered tokenized stocks in 2021 and early perp DEXs already existed, but outside demand and product quality were not yet ready.
Kyle is positioning for a selective altcoin cycle rather than requiring Bitcoin to make a new all-time high. With BTC around $84,000 during recording, he considered the approximately $58,000 bear-market bottom safely behind unless revisited; BTC could range around $90,000-$100,000 while strong alts keep grinding higher. His largest disclosed position was Venice, alongside Backpack and Collector Crypto, while Orbio was the higher-risk bet he thought might reach roughly $500 million—not necessarily $1 billion.
Deep dive
1. Concentration, not diversification, shaped Kyle’s investing apprenticeship
Kyle joined DeFiance Capital in 2025, after its 3AC-era crisis, and described watching Arthur Cheong invest as unusually instructive. Arthur’s defining trait was maximum conviction: where institutions commonly cap positions near 5%, he was comfortable putting 10%-20% into one name and letting it generate most of the return.
Kyle cautioned that he did not know the full 3AC history, but his understanding was that DeFiance’s original fund was clawed back before Arthur rebuilt through a second fund and made multiples. His takeaway was less about the episode than the temperament: “You know, kind of a beast.”
2. Hyperliquid proved that sustained growth can earn a sustained bid
Kyle’s starting premise was deliberately conventional: healthy capital markets require businesses whose underlying activity grows year after year. He compared that engine with the US economy, the S&P and the Mag Seven—especially Nvidia, which can be both among the world’s largest companies and among its fastest-growing.
Hyperliquid is crypto’s cleanest current specimen because its usage and economics produced the rare chart where, in Kyle’s words, “you DCA and you’ll be fine.” Capital follows the growth, price strengthens the ecosystem, and improving fundamentals attract still more capital—the same reflexive loop that concentrates flows in dominant public equities.
Hyperliquid’s roughly 99% buyback policy also reset competitive expectations. Three years earlier, buybacks were barely part of crypto’s vocabulary; now projects advertise monthly repurchases, while investor-relations tools and reports can make misalignment harder to hide. Kyle argued that these pressures are slowly manufacturing better fundamentals across the market.
3. Meme exhaustion sends capital back toward measurable reality
Kyle’s model of a meme cycle begins with one dominant coin, followed by progressively smaller imitators—perhaps $500 million, then $100 million, $30 million and $10 million. Because none has an underlying business, “there’s no ceiling to them, but there’s also no floor”; the same reflexivity that creates a vertical rally permits a 99% collapse.
Trump coin was his sharpest example: an unprecedented launch drove a near-immediate spike, followed by a roughly 15% decline the next day and continued deterioration toward about $2 by his recollection. Memes do not mechanically create fundamental investing, but saturation makes their distinctions impossible to defend, pushing investors toward revenue as an anchor.
DeFi Dad preserved the counterpoint: researchers such as Michael Nadeau have reportedly found on-chain meme signals and produced entries worth roughly 10X. Nomatic remained unconvinced by Murad-style holder size, distribution and retention metrics—“navigating in the dark” compared with revenue and TAM—but accepted that disciplined theses can exist inside the chaos.
The resulting market is K-shaped rather than universally fundamental. Crypto can still manufacture “fundamental-ish” narratives such as unused AI-agent projects, but Kyle expects about 90% of both memes and ostensibly serious investments to die; the 10% with genuine usage and revenue rise to the top.
4. The best DATs can distribute an altcoin, its utility and its yield
Kyle credited Evan for surfacing Stablecoin X near the bear-market bottom and admitted he was “so disappointed” in himself: he had independently imagined a winning DAT but “didn’t believe in crypto enough” to act sooner. He also declined to overstate expertise on Stablecoin X’s warrant mechanics without his notes.
His recurring chart begins with attention outrunning fundamentals, followed by an approximately 90% crash. If the project survives while its business continues growing, the second wave can recover and exceed the old high after everyone has written it off; Kyle said this pattern had made him more money than anything else.
Bitcoin’s DAT model is comparatively one-dimensional: issue equity or yield-like instruments, buy more BTC and increase Bitcoin per share. Altcoin DATs can do more—purchase the token, give traditional investors who cannot hold it exposure, and build new access or utility for the underlying network. Kyle pointed to PER, described as Hyperliquid’s debt, and said a fund associated with Stanley Druckenmiller was reported to have bought it for Hyperliquid exposure.
DeFi Dad framed the structure as “DeFi yield for the public markets,” contrasting it with borrowing simply to accumulate a non-yielding asset. Nomatic added an unverified but notable datapoint from a source he considered reliable: roughly $500 million allegedly wanted exposure to an unnamed token but could not buy it without a DAT.
5. “Getting it” matters more than beginning with the perfect idea
Nomatic said that, after seeing roughly 70-80 investments develop through their full cycles, he had returned to the view that the founder and team matter most because a promising idea may need to pivot or iterate five or ten times under adversity.
Kyle’s definition of founder quality was business sense: “the ability to convert your determination and your intellect and your skills into a tangible business and have the business sense to run it well.” The difficulty is recursive—“they have to get it, and you have to get it that they get it.”
He sees that quality in Ethena’s Guy Young and Hyperliquid’s Jeff Yan. Nomatic, by contrast, moved from being a Circle bull—attracted to a pure-play stablecoin exposure—to believing Jeremy Allaire does not “get it, get it,” leaving room for Ethena to challenge Circle through faster execution and yield sharing.
Nomatic criticized Coinbase and Base as insufficiently responsive, arguing that Vlad and Robinhood were moving faster. The hosts softened that verdict: Coinbase has fought unseen regulatory battles, enabled broad crypto access and executed the “DeFi mullet” well, even if Robinhood appears more attentive to users.
DeFi Dad offered Derive as the complementary case. Nick Forster kept building on-chain options before the category was fashionable, and Derive was taking share from Deribit as reported notional volume rose about 5X year over year—evidence that endurance and adaptation matter after the original thesis is formed.
6. Inference is the highest-torque point in Kyle’s AI thesis
Kyle’s top-down question is which part of the AI stack benefits most directly from expanding use. His answer is inference—the act of running a trained model to produce an answer—which he described as exhibiting Jevons paradox because demand is rapidly increasing.
He pointed to OpenRouter usage and Muse, which he said topped the charts and was downloaded more than ChatGPT over the same amount of time. Manus then appeared, and Kyle said Muse validated the conversational-agent model he had highlighted through “Grok Bot”: users do not want to configure agents; they want to ask for an outcome and have the software execute it.
Public-market investors can own Nvidia, memory, PCBs or underlying commodities, but those are indirect. Quoting Galaxy’s taxonomy, Kyle divided crypto exposure into owning inference access, mining through AI work and lending for GPUs; owning access is his favorite because it most directly captures a doubling in inference demand.
7. Venice is Kyle’s cleanest liquid proxy for AI usage
Private frontier-model companies are difficult to own, while Chinese vehicles including Zhipu, Alibaba and Tencent are, in Kyle’s view, imperfect expressions of the trade. Venice instead offers a user-facing model whose token is more directly associated with usage, making VVV his preferred way to “long inference.”
Staked VVV earns more VVV; staking 100 provides a pro subscription; and locked staked VVV creates DM. Each DM supplies $1 of Venice usage per day “forever” on any model, so greater demand for DM locks VVV, while revenue burns VVV through a mechanism Kyle said currently produces only modest effects.
Kyle initially dismissed Venice’s privacy positioning—“who cares”—but changed his mind as frontier models increasingly restricted conversations. He now sees privacy access as a real usage driver, reinforced by the hockey-stick growth chart founder Eric Voorhees shared.
Nomatic’s pushback was token-equity alignment: an equity raise coincided with a price dip because holders had assumed the token was the primary value-accrual instrument, yet the system also includes DM. His reluctant conclusion was that a trade can still work if the market broadly accepts it as “the AI trade,” even without every tokenholder-rights box checked.
8. Orbio is the high-beta end of the inference barbell
Kyle entered Orbio near an $8 million market cap, when the founder mattered more than a fully formed product. He relied partly on strong references around the builder and the founder’s connection to a team that had produced Elon, the founder of Pump Fund, and others: “This guy’s, like, a goated builder.”
The original product was a literal inference marketplace, allowing unused inference capacity to be sold to buyers seeking perhaps $10 of AI credits. Kyle liked it because rapidly expanding inference demand should create large users who want to buy it, while a genuinely novel marketplace can command a premium in crypto.
Orbio then added a launchpad where agents can purchase inference from the same marketplace, suggesting a reflexive loop: agents launch, consume credits, replenish through Orbio and continue operating. Kyle said he had not spoken to the founder, so presented that grand vision as a possible direction rather than a confirmed strategy.
His distilled thesis was “good, great founder, good vertical, novel product” plus an on-chain token below roughly $80 million. He floated $1 billion, immediately moderated it, and settled on approximately $500 million as a plausible hit comparable with Pond’s reported peak.
9. Outside demand separates durable sectors from crypto-native mirages
Kyle’s two-part filter requires product-market fit—meaning underlying growth—and an intersection with TradFi or another evergreen external market. Pure crypto revenue is too fickle: OpenSea can look like a unicorn one moment and go to zero the next when the internal cycle turns.
His S-tier consisted of inference, tokenized stocks and DATs, with privacy added despite his admitting, “I have zero Zcash.” Perp DEXs and stablecoins were A-tier. The unifying feature was net-new, steady-state demand entering crypto rather than existing capital merely rotating between narratives.
Timing is the crucial variable. Mirror offered mirrored Tesla exposure on the UST chain in 2021, while dYdX and GMX established perp DEXs early; today, banks are pushing tokenization, BlackRock is talking about it, Hyperliquid has steadily growing perp volumes and equity perps demonstrate broader demand.
Kyle’s indictment of the previous false start was memorable: the first DAT wave became “a bunch of crypto people dressed up in suits selling to other crypto people, thinking that they were the suits.” Regulation and institutional optimism arrived, but usable products and genuine outside buyers had not reached the same inflection point.
10. A K-shaped cycle favors patient ownership of a few compounding winners
Kyle described liquid investors as “spoiled for choice” after years when the fundamentals thesis was directionally right but chronically early. Backpack—reportedly up about 4X-5X from its bottom—stood out for a long-term-oriented founder; Collector Crypto remained one of his favorite businesses despite weak token performance because of its tokenized RWA, gacha and collectible vertical.
His disclosed core names were Venice, his largest position and one he wanted to increase; Backpack; and Collector Crypto. The common strategy was not constant rotation but finding projects whose numbers resemble Hyperliquid’s and holding them: chasing successive rallies may work repeatedly, “and then it dies.”
Nomatic agreed that opportunity had flipped from early-stage venture bets toward liquid tokens around late 2024 or early 2025. As the industry consolidates and fewer blue-ocean ideas remain, he and DeFi Dad have nearly stopped angel investing except for occasional outliers.
On the broader market, Kyle called October pivotal: it distinguishes an early-cycle setup, where the following months may be good, from a late-cycle setup, where they may be bad. He believed the bear-market bottom was behind, but his conviction was conditional—“you have to believe in something”—rather than a precise BTC, ETH, SOL or Zcash forecast.
With BTC around $84,000 during recording, Kyle’s practical invalidation reference was the prior bottom near $58,000. Bitcoin might never make a new all-time high and could simply range around $90,000-$100,000; barring a 5%-10% BTC day, he expects strong alts increasingly to trade on their own fundamentals.