AI power orders materialize; on-chain value capture continues to diverge | Five hard facts for October 7
① Wärtsilä secured an order for 282 MW of on-site power at a U.S. data center, including 15 50SG engines. The order has been included in Q3 2026 bookings, with delivery planned for 2028 and commissioning in 2029. Research implication: Constraints on AI infrastructure are extending beyond computing capacity to power that can be delivered on schedule. The order value, gross margin, and long-term service revenue have not yet been disclosed.
② On October 6, Pump used 10.5K SOL, worth approximately $1.28 million, to buy and burn 201 million PUMP tokens. In the preceding week, the protocol generated $16.32 million in fees and bought back and burned $7.6 million worth of tokens. Research implication: The link between revenue and actual token burns has been demonstrated, but the buyback policy can change, the token confers no legal claim to revenue, and future supply must still be assessed separately.
③ Aave V4 Core Hub added $22.83 million in USDT and lent out $21.20 million, bringing utilization to 92.8%; borrowing on the Ethena Spoke is nearing its previous cap. Research implication: Real borrowing demand has been validated, but evidence of updated funding for AAVE buybacks, on-chain authorization, and the first purchase order remains at 0/3. Protocol usage does not directly translate into returns for holders.
④ Bullish’s September trading volume was $48.6 billion, up 14.4% month over month and 22.7% year over year. The average trading spread was 2.14 bp, essentially unchanged from August. Research implication: Trading activity and pricing improved, but volume and spreads are not GAAP revenue and do not prove that the tokenized securities business has generated profits.
⑤ The U.S. 10-year real yield fell to 2.91%, down 4 bp on the day; COMEX Dec26 gold settled at $4,187.10 per ounce. Research implication: Discount-rate pressure eased marginally, but a single day’s yield move and futures settlement are insufficient to confirm a new long-term trend.
All information above is based on primary sources available as of October 7, 2026, 10:30 (UTC+8). Demand, trading activity, revenue, and the realization of ultimate value are not the same stage.
① Wärtsilä secured an order for 282 MW of on-site power at a U.S. data center, including 15 50SG engines. The order has been included in Q3 2026 bookings, with delivery planned for 2028 and commissioning in 2029. Research implication: Constraints on AI infrastructure are extending beyond computing capacity to power that can be delivered on schedule. The order value, gross margin, and long-term service revenue have not yet been disclosed.
② On October 6, Pump used 10.5K SOL, worth approximately $1.28 million, to buy and burn 201 million PUMP tokens. In the preceding week, the protocol generated $16.32 million in fees and bought back and burned $7.6 million worth of tokens. Research implication: The link between revenue and actual token burns has been demonstrated, but the buyback policy can change, the token confers no legal claim to revenue, and future supply must still be assessed separately.
③ Aave V4 Core Hub added $22.83 million in USDT and lent out $21.20 million, bringing utilization to 92.8%; borrowing on the Ethena Spoke is nearing its previous cap. Research implication: Real borrowing demand has been validated, but evidence of updated funding for AAVE buybacks, on-chain authorization, and the first purchase order remains at 0/3. Protocol usage does not directly translate into returns for holders.
④ Bullish’s September trading volume was $48.6 billion, up 14.4% month over month and 22.7% year over year. The average trading spread was 2.14 bp, essentially unchanged from August. Research implication: Trading activity and pricing improved, but volume and spreads are not GAAP revenue and do not prove that the tokenized securities business has generated profits.
⑤ The U.S. 10-year real yield fell to 2.91%, down 4 bp on the day; COMEX Dec26 gold settled at $4,187.10 per ounce. Research implication: Discount-rate pressure eased marginally, but a single day’s yield move and futures settlement are insufficient to confirm a new long-term trend.
All information above is based on primary sources available as of October 7, 2026, 10:30 (UTC+8). Demand, trading activity, revenue, and the realization of ultimate value are not the same stage.