1 Rising Cryptocurrency I'll Be Buying in September, and 1 I'll Definitely Avoid

Bittensor's TAO rose 21% over the 30 days through Sept. 10, 2026, as its August V440 upgrade redirected token emissions away from low-demand subnets toward more productive AI-service projects. The network still appears heavily subsidized, issuing roughly 3,600 TAO per day—about $316 million annually at $241 per token—versus estimated annual customer revenue of just $3 million to $15 million. Pump.fun gained 41% over the same period and burned more than $27 million of tokens in August, but competitive risks are rising after Robinhood Chain-based rival Pons generated $96 million in 30-day transaction fees, nearly twice Pump.fun's level.

Analysis

TAO’s investability hinges on whether the emissions reallocation creates genuine external-demand flywheels rather than simply concentrating token subsidies into a smaller set of validators and miners. The near-term effect is likely higher returns for winning subnets and a stronger staking narrative, but the valuation anchor remains weak until paid AI-service revenue grows materially faster than emissions. Over the next 1-3 months, subnet-token prices, active paying customers, and stablecoin/fiat-denominated service revenue—not aggregate network activity—are the relevant confirmation metrics; absent that evidence, TAO remains a high-beta AI/crypto liquidity trade.

PUMP’s buyback-and-burn model can support momentum during speculative bursts, but it is economically closer to a variable-volume transaction-fee annuity with minimal switching costs than a durable exchange franchise. Competition shifts the risk from token dilution to fee compression: a rival can subsidize launches, offer distribution through an existing brokerage user base, or reduce fees, forcing PUMP either to sacrifice take rate or allow volume to migrate. HOOD has optionality if its chain converts crypto engagement into funded accounts, trading balances, and cross-sell, but chain-level fee activity alone should not be capitalized into HOOD earnings until management discloses ownership economics, compliance costs, and revenue share.

The contrarian risk is that the apparent TAO improvement is already a widely understood supply-side catalyst while the hard problem—enterprise-grade reliability, customer acquisition, and defensible AI-service differentiation—takes 6-18 months to solve. Conversely, PUMP may be underappreciated as a tactical short-term beneficiary if retail risk appetite persists, because burns mechanically amplify per-token optics; that is not a reason to underwrite a multi-quarter long. A broad crypto drawdown, Solana congestion or outage, or regulatory restrictions on retail token issuance would impair both assets, with PUMP more exposed to activity velocity and TAO more exposed to liquidity and AI-narrative multiple compression.

Key Decisions for Investors

• Treat TAO as a tactical, size-limited long only after verifying two consecutive monthly increases in externally paid subnet revenue and active customers; use a 1-3 month horizon and exit if token emissions remain the dominant source of miner economics or TAO underperforms SOL by more than 15% after the next network-data release.
• Avoid strategic PUMP exposure; for liquid venues, consider a 1-3 month relative-value short PUMP versus long SOL only if PUMP’s fee share continues declining while total Solana launch activity remains stable. The thesis is invalidated if PUMP preserves share and buybacks exceed net new token supply for two consecutive months.
• Keep HOOD on an event-driven watchlist rather than adding on third-party chain-fee estimates. Upgrade only if quarterly disclosures show Robinhood Chain driving measurable crypto trading revenue, funded-account growth, or materially lower customer-acquisition cost; otherwise, its stock remains primarily exposed to retail trading volumes, rates, and broader risk appetite rather than launchpad economics.
• Use NVDA/AI infrastructure exposure rather than TAO as the core AI allocation: TAO offers convex narrative beta but lacks auditable revenue quality, whereas listed AI infrastructure has clearer demand visibility. Any TAO allocation should be funded from the speculative digital-asset sleeve, not by reducing core semiconductor exposure.