Around September 13, a post from Serenity (X account @aleabitoreddit, "White-haired Stock God") quickly spread in the Chinese community, focusing not on the revenue figures in Oracle's Q1 financial report, but on a line buried in the conference call details—a footnote stating that all GPU capacities entering the renewal stage were renewed or resold, with transaction prices increased by about 20% compared to the original contracts. More counterintuitive is that these GPUs were not new arrivals, but mostly "old" equipment that had been in service for about four years. According to traditional hardware depreciation logic, their renewal prices should ideally be discounted for clearance, but in reality, they experienced price increases upon renewal. This sharp rise in the value of old GPUs directly counters Michael Burry's earlier assertion that "GPU computing assets will rapidly depreciate due to technological iterations," and has thus been interpreted by Serenity and some Chinese media as a potential positive signal for the computing power sector—not just for companies like Nebius (NBIS), IREN (IREN), but also for Nvidia (NVDA) and the broader Neocloud sector, triggering a new round of debate about whether GPUs are ultimately short-lived consumables or long-term cash cows.
Four-Year-Old GPUs Renewed Yet Price Increased by 20%
To understand this debate about whether "GPUs are short-lived consumables or long-term cash cows," one must return to the numbers Oracle provided. According to their latest quarterly financial report and conference call disclosures, the GPU capacity entering the renewal stage has not been widely abandoned or left idle; instead, they were either renewed by the original customers or resold to new clients, achieving "full absorption." More critically, a significant portion of these computing assets have been in service for about four years or longer, yet they secured prices approximately 20% higher than the original contracts during renewal or resale—meaning the same old card is actually more expensive in the second contract than in the first.
This phenomenon creates a stark conflict with traditional hardware depreciation logic. Common sense dictates that after four years of use, servers and GPUs should have completed depreciation, with renewal prices being discounted or pushed out by newer products. However, Oracle's financial report indicates the opposite: the renewal unit price of older GPUs has risen by an average of about 20%. The issue lies in the fact that public information is only surface-level, stating "all capacities were renewed or resold, and prices were adjusted upward," without breaking down specific models, quantities, unit prices, total contract amounts, and renewal durations. Additionally, there is no detailed explanation of customer structure, leaving external investors to make secondary inferences based on this summary. Therefore, this data can only be viewed as a strong yet still rough signal: in the current supply-demand environment, at least some of the four-year-old GPUs in Oracle's hands still possess the ability to renew at a premium, but this is insufficient to support a unified conclusion about the lifespan and pricing elasticity of all computing assets.
Big Short Prophecy Fails? The Depreciation Argument Faces Backlash
In traditional financial narratives, Michael Burry's name has almost become synonymous with "bubble identifier": he precisely shorted the U.S. subprime mortgage market, granting him exceptional discourse power on asset valuation and risk pricing. Extending into the era of computing power, his proposed framework is not complex—GPU computing fundamentally remains a hardware asset, with the speed of technological iteration far surpassing that of industrial era machinery. Rapid upgrades compress the economic lifespan of older cards. If the next generation of GPUs overall crushes the previous generation in efficiency, bandwidth, and instruction sets, then the utility value and bargaining power of old cards should decline linearly or even accelerate year by year. Depreciation becomes more than just an accounting number; it's a cash flow loss: rent reductions, lowered utilization rates, ultimately manifesting as "rapid depreciation" on the balance sheet.
Because of this underlying logic, the set of renewal data in Oracle's most recent financial report stands out: the majority of GPUs that have been used for about four years have not been abandoned or depreciated during the renewal phase; instead, they were all successfully renewed or resold, and in their new contracts, they secured an approximately 20% price premium over the original contracts. According to traditional hardware depreciation intuition, this almost represents an "inverse trajectory" of asset pricing, directly conflicting with the notion that "technological iterations will inevitably lower the prices of old GPUs quickly." Serenity seized this point and interpreted it in the Chinese community as a real refutation of Burry's argument on "rapid GPU depreciation," with some media even using "the big short gets hit in the face" as a narrative hook, amplifying this cross-market collision of viewpoints. However, returning calmly to the data itself, it remains only the result of a single large cloud vendor in a specific supply-demand environment. In the absence of similar renewal and pricing information from other cloud service providers and the broader GPU market, this counterexample is insufficient to fundamentally overturn the overall logic of "GPU computing assets facing high technological depreciation risk;" it can only be viewed as an important corrective signal to overly simplistic narratives about "rapid depreciation."
Computing Power Renters: Old Inventory Becomes Gold
For companies like Nebius and IREN, which sell "computing power rentals" based on GPUs as the underlying assets, the detail extracted by Serenity from Oracle's financial report most directly impacts their business model. In traditional financial models, GPUs that have been in use for about four years are often seen as old equipment that should have their renewal prices lowered. But Oracle's reality is that this old batch of cards was not only fully renewed or resold, but the renewal price was even approximately 20% higher than the original contract. Based on this, Serenity determines that if a four-year GPU can still fetch rent higher than the original contract, then the inventory of those GPUs already in the "second half of depreciation" at Nebius and IREN need not wait for scrap value, but instead may become cash flow assets capable of passive price increases and extended charging periods.
Once the market accepts the premise that "old cards can raise rent," the entire narrative for computing power renters will be rewritten: the economic lifespan of the same batch of GPUs will no longer be crudely drawn as a straight line of three to four years, but may maintain monetization potential for a longer time. Coupled with the price increase upon renewal, this means that the asset recovery rate for each piece of hardware improves and future cash flow curves become smoother, theoretically leaving room for the long-term cash flow stability and valuation re-evaluation of models like Nebius and IREN. However, as of now, these potential benefits remain at the speculative level posited by Serenity and some media, with neither Nebius nor IREN releasing any official response regarding Oracle's data, nor have more operating or pricing details been disclosed. Whether the inventory of old cards can similarly "turn to gold" across different companies and under different supply-demand conditions remains an open question that requires subsequent empirical validation.
From Old Cards to New Cards: Dual Pull of Computing Power Demand
If Nebius and IREN are considered the "direct beneficiaries" of Oracle's old GPU premium, what excites the market even more is the potential upside for Nvidia and the Neocloud sector, which primarily offers cloud-based GPU computing power services. The logic of interpretation isn't complex: old cards that enter the renewal stage and have been used for over four years not only haven’t been depreciated, but also overall rose in price by about 20% during renewal or resale, with all capacities fully absorbed. This serves as real transaction data telling everyone—the gap in computing power is still large enough at this stage to overturn the traditional hardware depreciation model. For upstream suppliers like Nvidia, downstream cloud vendors willing to pay higher prices for "old cards that should be discounted" often means that overall demand for computing power is tight, raising the pricing power and sales expectations for the new generation of GPUs; for the Neocloud sector, the signal that "old cards can earn money longer" is seen by many as providing an extra time dimension endorsement for its cloud computing power business model.
More intriguingly, this is not a simple linear narrative of "old cards squeezing out new cards," but a dual pull dynamic potentially coexisting with the extension of old GPU economic lifespans and the expansion of demand for new cards: on one side, four-year-old devices are still raising prices during renewals, extending their profit cycles; on the other hand, overall tight demand for computing power allows stronger bargaining power for higher-performance new generation chips. This state of supply and demand drags both Nvidia and Neocloud into the same expectation curve. However, it must be emphasized that Oracle's Q1 financial report remains an isolated sample; the view that the price increase for old GPU renewals is beneficial to Nvidia and Neocloud is strictly based on Serenity and some media's extrapolation from a single data point, lacking support from Nvidia or Neocloud's own operational data and public responses. Therefore, before more companies present their renewal data and pricing curves, whether Nvidia and Neocloud can continuously gain fundamental benefits from this "old card premium" remains an open market hypothesis.
What Can Oracle's Financial Report Show?
The approximately 20% increase in the renewal price of Oracle's old GPUs brings Michael Burry's assertion that "GPU computing assets will rapidly depreciate due to technological iterations" back to the realities of supply and demand: during a phase of extreme computing power scarcity, hardware no longer simply falls along textbook depreciation curves; old cards over four years can still secure premiums in renewals, and economic lifespans are redefined by demand. Serenity extends this clue to sectors such as Nebius, IREN, Nvidia, and Neocloud, and the market is eager to use it to perform cross-company and cross-industry valuation reassessments, as if one financial report has rewritten the pricing logic of the entire computing asset landscape. However, looking at it calmly, we currently have only Oracle as a single large cloud service provider's Q1 renewal data, lacking similar samples from more cloud vendors and computing power operators. All conclusions regarding the longevity of old GPUs and the elevation of revenue curves are merely forward-looking speculations based on limited information. At this point, it is more worth paying attention to whether other cloud vendors' GPU renewal prices, utilization rates, and service years change directionally in the upcoming financial report season, using more comprehensive evidence to determine whether this "old card premium" is merely random noise or indicative of a structural change in the pricing approach for computing assets.
Join our community to discuss and grow stronger together!
Exclusive Hyperliquid benefits for AiCoin: https://app.hyperliquid.xyz/join/AICOIN88
Exclusive Aster benefits for AiCoin: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin's on-chain Twitter: https://x.com/aicoinwhaledata
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
