gpu

Fully Homomorphic Encryption (FHE) company Zama CEO Rand Hindi stated on X on Monday that the company has achieved benchmark results of 1,000 confidential transactions per second on GPUs, an increase of about 5,000 times compared to the 0.2 transactions per second demonstrated at ETHcc in 2022. Hindi mentioned that the cost of running an FHE node is approximately $0.000004 per ERC7984 transaction, significantly lower than zero-knowledge proofs, but this comparison has not been independently verified. The benchmark has not yet been activated on any mainnet, is self-reported data from the company, and the testing methods have not been published, with plans to go live by the end of the year.Hindi also set long-term goals: 10,000 tps by 2028 and 100,000 tps by 2030. Zama has been active recently: on July 24, a confidential RFQ exchange product was launched, which within weeks shielded over $121 million USDT on Ethereum, collaborated with Elliptic to enhance compliance screening for confidential transactions, and co-founded the Confidential Token Association with OpenZeppelin and Inco to standardize encrypted ERC-20 tokens. Zama previously completed a $57 million Series B funding round.

According to a report by the WSJ, Nvidia has invested in the AI laboratory Safe Superintelligence (SSI), founded by former OpenAI chief scientist Ilya Sutskever. The two parties have reached a long-term cooperation agreement aimed at expanding SSI's computing resources while helping Nvidia secure important clients in the AI field. Both companies stated that Nvidia made a "large-scale" investment after understanding some of SSI's research progress, but did not disclose the specific amount.As part of the collaboration, SSI will receive a significant amount of Nvidia's flagship GPU resources, with its computing power expected to increase by an order of magnitude. Previously, SSI primarily relied on TPU chips provided by Google for AI research and development. This collaboration indicates that Nvidia is further expanding its AI chip ecosystem and binding future computing power demands through investments in top AI laboratories.SSI was established in 2024 by Ilya Sutskever, with the goal of developing "Safe Superintelligence." The company has previously raised about $2 billion in funding, with investors including Andreessen Horowitz and Sequoia Capital, and reached a valuation of approximately $30 billion last year. This is not Nvidia's first investment in an AI company founded by former core members of OpenAI. In March of this year, Nvidia also invested in Thinking Machines Lab, founded by former OpenAI chief technology officer Mira Murati, whose first AI model is trained on Nvidia hardware.Ilya Sutskever is one of the important researchers in the field of modern artificial intelligence, having co-founded OpenAI and promoted the development of large model technologies such as ChatGPT. However, after leaving OpenAI, he began to question the approach of solely relying on expanding data and computing power to drive AI progress, turning instead to explore new directions in superintelligence research.

Libbitcoin developer @evoskuil stated that the latest version of Libbitcoin Server can complete the initial block download of a Bitcoin full node in 55 minutes and perform full verification. This speed improvement comes from a new GPU-based signature verification system that runs in parallel with conventional script verification, rather than as a separate step.The software does not pause execution to verify each ECDSA and Schnorr signature; instead, it captures signature checks, batches them, and processes them via GPU while the remaining verification continues. @evoskuil indicated that this method achieves about a 99.5% signature capture rate without blocking verification, thus eliminating the need for a second verification. If a batch signature check fails, the software traces it back to the corresponding block and marks that block as invalid. He also mentioned that this design can batch process an almost arbitrary number of signatures, keeping the GPU ahead of the remaining verification process and significantly shortening the full node synchronization time.

According to official news, Kimi stated that in the past 48 hours, the demand for Kimi K3 has exceeded expectations, and GPU resources are nearing the current capacity limit. To ensure the experience of existing subscription users, the platform has temporarily suspended new subscriptions, and current subscribed users will not be affected. Kimi mentioned that it is working to expand computing power as soon as possible and will gradually open new subscription slots in batches.Meanwhile, the platform will split the membership system into two types of plans in the future: one for Kimi members on the Kimi web portal, App, and Work, and the other for Kimi Code members focused on programming workflows, to more accurately match computing resources and maintain service stability.

According to The Block, Bernstein pointed out in its latest research report that the perpetual contract structure originating from the cryptocurrency market has first entered the AI computing power market. The Architect AX exchange, registered in Bermuda and not regulated by the CFTC, has launched GPU computing power perpetual contracts, which are anchored to spot prices through funding rates, a mechanism that directly originates from the cryptocurrency market. The cash-settled computing power futures from CME and ICE are still awaiting CFTC review, with a launch expected by the end of 2026.In addition, the CFTC-regulated prediction market platform Kalshi has previously launched prediction contracts linked to GPU rental prices and introduced the computing power pricing curves for B200, H200, and A100 chips on July 14, with the B200 quoted at $5.41 and a historical peak of $7.39.Bernstein noted that unused GPU computing power cannot be stored, and hedging is the only way to bear price risk. However, the current market liquidity is still dominated by speculative funds, and the difficulty in constructing computing power benchmark indices is also a major bottleneck in market development.

According to BBX data, yesterday the global US stock market and companies listed on the Tokyo Stock Exchange made significant moves in the accumulation of digital asset reserves. The latest updates from two core companies are as follows:GPUS core treasury breaks the 1,000 mark: Artificial intelligence data center company Hyperscale Data, Inc. (NYSE American: $GPUS) officially announced that the number of bitcoins held in its treasury has officially exceeded 1,000. Management stated that the company will continue to closely align with market conditions and existing capital allocation opportunities, supported by its core data center business, to continuously manage and expand its bitcoin reserve scale.Metaplanet raises $137 million through targeted private placement: Metaplanet Inc. (TSE: $3350) announced that it will issue new shares to specific institutional investors through the Third-Party Allotment mechanism under Japanese securities law, aiming to raise approximately $137 million (equivalent to 21.4 billion yen), with all funds to be used for purchasing bitcoin. The company accumulated 2,823 BTC in Q2 this year, and as of June 30, its total holdings have reached 43,000 BTC (valued at approximately $2.67 billion). Based on the current market price of about $62,000 per coin, this lightning private placement is expected to add approximately 2,210 BTC to its reserves, helping it strive towards its ultimate goal of 100,000 BTC by the end of the year.

According to reports from Jiemian News, Kingsoft Cloud will accelerate the construction of GPU computing clusters in the second half of the year to meet the explosive growth in computing power demand from major clients. Among them, Xiaomi's demand for Kingsoft Cloud's GPU computing power has upgraded from a ten-thousand-card cluster, with the related budget significantly increasing from nearly 4 billion yuan to over 10 billion yuan. In addition, Alibaba's large model team has signed a 5-year computing power leasing contract with Kingsoft Cloud, involving more than 3,000 eight-card GPU servers. Based on the contracted price, the annualized revenue after full delivery will exceed 4 billion yuan.To meet the surging customer demand, Kingsoft Cloud's capital expenditure plan for 2026 has been raised to 15 billion yuan, with an annual revenue target of 12.5 billion to 13.5 billion yuan. It is reported that due to tight upstream supply, Kingsoft Cloud is currently only accepting long-term contracts from clients for 3 to 5 years, and some orders are facing delivery delays. Due to concerns about the risk of impairment of high-level card assets, Kingsoft Cloud is currently pausing aggressive hardware expansion, anticipating that the prices of computing hardware may reach a turning point for decline in the third quarter of this year.

According to a report by the Financial Associated Press, Nvidia announced a new AI infrastructure collaboration model on July 1 local time. Through revenue-sharing and credit-support mechanisms, it will jointly build large-scale, multi-tenant AI factories with AI cloud service providers, helping startups, model developers, enterprises, and research institutions to access AI computing power more quickly, while creating a sustainable revenue stream for Nvidia linked to computing power usage.According to the plan, AI cloud vendors will deploy AI factories based on the Nvidia DSX AI Factory architecture to provide cloud computing services. The first batch of partners includes Sharon AI and Firmus. Among them, Sharon AI plans to deploy up to 40,000 NVIDIA Grace Blackwell GB300 GPUs; Firmus will build a DSX AI factory park on Batam Island, Indonesia, with plans to expand to a power scale of 360 megawatts and deploy up to 170,000 Nvidia GPUs.

According to DataCenterDynamics, Bitcoin mining company and data center Hive Digital Technologies announced that it will raise $100 million by issuing zero-coupon convertible senior notes maturing in 2031 to further expand its computing power infrastructure.In addition, the company has signed a non-binding letter of intent (LOI) with a Swedish sovereign technology company that has an investment-grade credit rating, intending to lease approximately 30MW of computing power resources at its data center for a term of up to 10 years.According to the agreement, Hive will deploy approximately 25MW of IT computing capacity and plans to use up to 10,000 Nvidia GB300 GPUs; however, the letter of intent still requires further negotiation and the signing of a final formal agreement.

According to The Information, cloud computing startup Runpod announced the completion of a $100 million financing round and stated that it has rejected multiple acquisition offers, with specific investor information not yet disclosed.The company focuses on providing GPU computing power rental services for developers, supporting the deployment of open-source models and AI applications. Currently, this sector generally relies on NVIDIA GPU servers for computing resources. As AI applications expand, the market demand for low-cost, highly flexible computing infrastructure continues to rise, driving up the valuations and financing enthusiasm for cloud service providers like Runpod.

According to The Information, informed sources reveal that Goldman Sachs and JPMorgan are exploring trading methods based on the cost of computing power, including futures contracts related to GPU leasing prices.As one of the scarcest resources in the AI boom, related futures are expected to be listed by exchanges later this year. Industry insiders say this initiative demonstrates that hundreds of billions of dollars flowing into data centers and the chip sector are reshaping the financial market landscape. For banks providing financing for AI infrastructure construction, such innovative tools may become new means of managing risk.

The prediction market platform Polymarket has disclosed that it has completed its first institutional-level on-chain block trade, a six-figure dollar transaction executed by digital asset broker FalconX and AI risk clearing infrastructure startup Anera Labs on the Polygon blockchain. The transaction involved contracts related to the Ornn Compute Price Index, which is used to track the rental prices of NVIDIA H100 GPU chips.Polymarket stated that this is the first on-chain block trade involving institutional participation in the prediction market space. In the future, FalconX will also serve as the exclusive market maker for block trades on the Polymarket platform to support the liquidity needs and price discovery for institutional clients.

Bitcoin mining company IREN announced the completion of an investment-grade GPU financing totaling $3.65 billion to support its AI cloud contract signed with Microsoft.The financing includes $2.1 billion in U.S. private placement bonds and $1.55 billion in delayed draw term loans, with a combined debt cost of 6%.It is reported that this financing received an A rating from Fitch and an A (low) rating from DBRS, making it the highest-rated investment-grade GPU financing project currently publicly disclosed, and also the first GPU financing case in the U.S. private placement market.IREN stated that the financing is secured by GPU assets and related contract cash flows, which helps optimize the capital structure and supports the company in expanding AI Cloud capacity to 480MW by the end of 2026.

Morgan Stanley pointed out that by 2030, the global semiconductor industry market size could reach $1.5 trillion, with AI-related semiconductor products contributing a significant share. The capital expenditure of major cloud service providers remains strong.Morgan Stanley's cloud capital expenditure tracker estimates that by 2026, cloud capital expenditure will approach $811 billion. Research suggests that agent-based artificial intelligence has generated a growing opportunity for CPU applications. As AI shifts from reasoning to execution, GPU computing intensity increases. The agency has raised the total addressable market (TAM) for orchestration CPU in the baseline scenario to $79 billion, with the market added value (TAM) forecast for CPU orchestration technology reaching $238 billion.

According to Bloomberg, the parent company of the New York Stock Exchange, Intercontinental Exchange (ICE), announced plans to launch futures contracts based on "computing power" to track changes in the costs of computing power such as GPUs that support the AI industry. This product still requires regulatory approval.It is reported that ICE will collaborate with financial infrastructure company Ornn to develop a pricing system for related derivatives based on its GPU cost index. This index will be used to support computing power futures contracts.