Hyperliquid’s Fee Flywheel Pulls $172M Into HYPE ETFs as Token Sets New High While Bitcoin ETFs Bleed

The cleanest signal in a choppy quarter is divergence: while U.S. spot Bitcoin ETFs have seen nearly $5.6 billion in net outflows since mid-May, three new Hyperliquid ETFs have absorbed almost $172 million. The market is voting for cash flows over narratives.

HYPE, the protocol’s native token, has tracked that conviction. It’s up roughly 73% over the past month and 196% in 2026, touching a fresh all-time high of $75.96 on Tuesday, per CoinGecko. Cumulative trading volume across the trio of ETFs is approaching $900 million, with Bitwise’s BHYP leading on about $106.6 million in net inflows and $122.8 million in net assets, followed by 21Shares’ THYP at $60 million and Grayscale’s HYPG at $8.6 million, according to SoSoValue.

The center of gravity here is a simple but potent design: Hyperliquid’s fee-to-buyback loop. Between 97% and 99% of protocol trading fees are automatically routed to the Assistance Fund (AF), which buys back HYPE daily. That creates an immediate, non-speculative linkage between activity and token demand. Layer on the stablecoin stack: Coinbase serves as the official treasury deployer for existing USDC reserves, and the AQAv2 program allows $5 billion of USDC to earn roughly 4%—with 90% of that yield also redirected to the AF. The result is a compounding engine that scales with both trading volume and idle capital, not just market direction.

Institutions appear to be treating HYPE less like a commodity proxy and more like a high-growth, fee-generating platform. The psychology tracks: investors often pay up for mechanisms they can underwrite—visible revenue capture, recurring buybacks, and diversified lines of business. Hyperliquid’s expansion beyond crypto perpetuals into commodities, equities, outcome markets, and pre-IPO exposures gives that thesis breadth. The CBRS perpetual’s pre-market pricing landed within 1.3% of the stock’s eventual NASDAQ open, and SpaceX’s SPCX perpetual—listed via the permissionless HIP-3 framework—drew about $1.4 billion in volume in a single session, roughly 30% of that day’s HIP-3 activity, per hl.eco. That kind of product-market fit turns the AF from a nice idea into a measurable funnel.

Two things are doing the heavy lifting: - A deterministic buyback policy (97%-99% of fees to AF) that tightens float as volumes rise. - A yield-enhanced treasury (AQAv2’s 4% on $5 billion USDC, with 90% sent to AF) that deepens liquidity and reinforces the loop even in sideways markets.

This is why HYPE inflows have held up as Bitcoin ETFs faced macro headwinds like higher Treasury yields and geopolitical risk. When volatility increases, derivatives venues often see more hedging and repositioning, which keeps fee generation resilient; that stability is precisely what many allocators want from an “infrastructure” bet in crypto.

My take: the flywheel is the story. The market is rewarding protocols where unit economics are transparent and self-reinforcing. That said, durability hinges on a few execution variables. First, sustaining share gains in derivatives—user growth, liquidity depth, and reliable matching—matters more than headline price targets. Second, the treasury design benefits from scale but concentrates operational trust in a single deployer; that concentration should be actively stress-tested. Third, buyback-first models can overshoot in both directions—exuberance when volumes spike, reflexivity on drawdowns—so pacing and disclosures will influence institutional comfort.

Near term, watch participation breadth rather than price: unique traders, open interest dispersion, and cross-market volumes on HIP-3. The ETF tape is supportive for now—HYPE products have collected $172 million in a month while Bitcoin ETFs gave up billions—and options markets are assigning roughly a 10%-15% probability that HYPE tags $100 by the end of July, per Derive’s Nick Forster. If execution continues and revenue diversity holds, the fee flywheel remains intact—and that’s the part institutions can model.