WhiteLine Daily from the WuBlockchain team focused its latest edition on long-duration commitments forming around AI infrastructure. The report covered higher long-term targets from Marvell, a multi-decade nuclear power arrangement signed by Google, a reported debt package being explored by SpaceX for Nvidia AI chips, and a sharp jump in Lambda's backlog ahead of a planned IPO.
Marvell raises long-term revenue targets
At its Investor Day, Marvell raised its FY2028 revenue target to about $20 billion, above Wall Street expectations of roughly $18.2 billion. The company also projected FY2031 revenue of $70 billion to $90 billion, with a midpoint of about $80 billion, compared with current market expectations of around $46.85 billion.
Marvell also lifted its FY2029 custom chip revenue target to $12 billion from a previous $10 billion. WhiteLine Daily said custom chips have become a key part of Marvell's growth outlook as large cloud companies including Google step up investment in in-house AI chips.
The report said the midpoint of Marvell's FY2031 target sits about 71% above current market expectations, pointing to a company view on the future ramp of existing custom chip programs that is far above current Wall Street models.
Google signs long-term deal with Constellation
Google signed a long-term power partnership with U.S. nuclear operator Constellation, including a 20-year power purchase agreement. Under the arrangement, Constellation will invest more than $4.3 billion in 11 existing nuclear units in Illinois, Pennsylvania and New Jersey. The upgrades, which include turbines, steam generators and digital control systems, are expected to add about 890 MW of nuclear capacity.
The two sides also signed a 15-year agreement under which Constellation will provide another 2.7 GW of existing power in the PJM market. That brings the total power supply covered by the broader partnership to about 3.59 GW, though the actual newly added generation capacity is 890 MW.
Google and Constellation will also expand technical cooperation using Google Cloud and Gemini to improve grid and power plant operations.
WhiteLine Daily's takeaway was that Google used a 20-year power purchase contract to secure 890 MW of new nuclear capacity, locking in long-term demand first and then enabling Constellation to spend more than $4.3 billion to expand existing units, rather than simply signing for already available electricity.
Financial Times says SpaceX is seeking $40 billion for Nvidia AI chips
According to the Financial Times, SpaceX is seeking about $40 billion in financing led by Apollo Global Management to purchase Nvidia AI chips.
The proposed package includes about $10 billion in bank loans and $30 billion in investment-grade debt. Large bond investors including Pimco could also take part, and the deal is expected to be completed in 2027.
WhiteLine Daily noted that Elon Musk has previously said SpaceX data centers will continue using Nvidia chips and that the company plans to expand its AI infrastructure spending. If completed, the transaction would rank among the largest debt financings tied to AI chip purchases so far.
SpaceX, Apollo and Nvidia had not commented on the report at the time of publication.
The report's reading was that the proposed $40 billion GPU purchase would be financed entirely through loans and bonds, turning AI chip expansion directly into long-term demand for credit markets.
Lambda seeks up to $4 billion before IPO
According to the Wall Street Journal, Nvidia-backed AI cloud provider Lambda is in what it describes as its final funding round before a planned IPO, seeking up to $4 billion led by Blackstone and Coatue Management at an implied pre-money valuation of about $14.5 billion. The company plans to go public in 2027.
Lambda's backlog rose from about $15 billion in June to about $50 billion in September, the report said. But roughly $35 billion of that increase came mainly from a single long-term compute contract signed in August by Anthropic for $35 billion, meaning the surge in backlog was highly concentrated in one large customer.
WhiteLine Daily said Lambda added $35 billion of backlog in three months, but almost all of it came from one order of the same size from Anthropic. In that reading, the customer concentration behind the $50 billion figure stands out just as much as the growth itself.
Main thread of the day
The report summed up the day's main thread this way: AI expansion is taking the form of different kinds of long-term commitments. Marvell is offering revenue targets far above market expectations, Google has signed a 20-year power purchase agreement, SpaceX is seeking $40 billion in debt financing, and Lambda has pushed long-term compute backlog to $50 billion. What comes next, the report said, is whether those commitments can be converted into revenue, power supply and cash flow on schedule.