7*24

Hot topic

Market

Activities

On July 14, Morgan Stanley released a latest research report, raising its capital expenditure forecasts for five hyperscale cloud providers—Meta, Amazon, Microsoft, Google-C, and SpaceX—for 2027/2028 by 9%/10% respectively, to approximately $1.2 trillion and $1.4 trillion. The main drivers for the upward revision include a roughly 20% increase in GPU costs, a lengthening of data center construction cycles to three years, and computing power expected to grow from about 30 GW in 2025 to around 120 GW in 2028. Meanwhile, the potential revenue space for Meta, Amazon, and others is far from being priced in by the market.

[Morgan Stanley: Top Five Cloud Firms' CapEx to Reach $1.2 Trillion Next Year]On July 14, Morgan Stanley released a latest research report, raising its capital expenditure forecasts for five hyperscale cloud providers—Meta, Amazon, Microsoft, Google-C, and SpaceX—for 2027/2028 by 9%/10% respectively, to approximately $1.2 trillion and $1.4 trillion. The main drivers for the upward revision include a roughly 20% increase in GPU costs, a lengthening of data center construction cycles to three years, and computing power expected to grow from about 30 GW in 2025 to around 120 GW in 2028. Meanwhile, the potential revenue space for Meta, Amazon, and others is far from being priced in by the market.

Morgan Stanley: Top Five Cloud Firms' CapEx to Reach $1.2 Trillion Next Year

According to Beating's monitoring, Tencent Hunyuan has released 1-bit and 4-bit compressed versions of Hy3, significantly lowering the deployment threshold for this 295-billion-parameter model. The original Hy3 weights are close to 600GB. The 1-bit version shrinks to 85.5GiB and can run on a single 96GB NVIDIA H20 GPU. When running on a single GPU, users need to disable some acceleration features and shorten the text length processed at one time. With two H20 GPUs, acceleration can be enabled. The 4-bit version has a volume of 169.9GiB and requires at least two H20 GPUs. It consumes more VRAM but yields results closer to the original model. Tencent Hunyuan states that the 4-bit version performs close to the original model in code, tool invocation, long text comprehension, and Agent tasks. The 1-bit version shows only a slight decline. With generation acceleration enabled, the 1-bit version speeds up by about 50%, and the 4-bit version by nearly 60%. Tencent Hunyuan also provides a 4-bit version suitable for server deployment, which can be made into an API for multiple users or applications to call.

[Tencent Hunyuan Squeezes Hy3 to 1-bit, 295-Billion-Parameter Model Runs on Single GPU]According to Beating's monitoring, Tencent Hunyuan has released 1-bit and 4-bit compressed versions of Hy3, significantly lowering the deployment threshold for this 295-billion-parameter model. The original Hy3 weights are close to 600GB. The 1-bit version shrinks to 85.5GiB and can run on a single 96GB NVIDIA H20 GPU. When running on a single GPU, users need to disable some acceleration features and shorten the text length processed at one time. With two H20 GPUs, acceleration can be enabled. The 4-bit version has a volume of 169.9GiB and requires at least two H20 GPUs. It consumes more VRAM but yields results closer to the original model. Tencent Hunyuan states that the 4-bit version performs close to the original model in code, tool invocation, long text comprehension, and Agent tasks. The 1-bit version shows only a slight decline. With generation acceleration enabled, the 1-bit version speeds up by about 50%, and the 4-bit version by nearly 60%. Tencent Hunyuan also provides a 4-bit version suitable for server deployment, which can be made into an API for multiple users or applications to call.

Tencent Hunyuan Squeezes Hy3 to 1-bit, 295-Billion-Parameter Model Runs on Single GPU

On July 13, SemiAnalysis memory analyst Ray Wang stated in an interview with CNBC that CXMT is expected to capture approximately 10% of the global DRAM market this year, making it a competitor that cannot be ignored. Although its conventional memory technology still lags behind Samsung and SK Hynix by about three years, and there is a three-to-four-year gap in the HBM field, under the growth of AI demand and overseas HBM export restrictions, CXMT is poised to continue expanding its market share in China. Its IPO will also become the most anticipated project in China's semiconductor sector this year.

[Video | SemiAnalysis Memory Analyst: CXMT to Be China's Hottest Semiconductor IPO This Year]On July 13, SemiAnalysis memory analyst Ray Wang stated in an interview with CNBC that CXMT is expected to capture approximately 10% of the global DRAM market this year, making it a competitor that cannot be ignored. Although its conventional memory technology still lags behind Samsung and SK Hynix by about three years, and there is a three-to-four-year gap in the HBM field, under the growth of AI demand and overseas HBM export restrictions, CXMT is poised to continue expanding its market share in China. Its IPO will also become the most anticipated project in China's semiconductor sector this year.

Video | SemiAnalysis Memory Analyst: CXMT to Be China's Hottest Semiconductor IPO This Year

On July 14, the Financial Times reported, citing insiders, that DeepSeek has begun approaching new investors. The company is considering launching a second round of financing, targeting a pre-money valuation of approximately $71 billion. The plan is still in preliminary discussions, and DeepSeek has not yet responded. DeepSeek completed its first round of funding around the end of May, raising about $7 billion, with a post-money valuation of approximately $52 billion. Liang Wenfeng personally contributed around $3 billion. Tencent, CATL, JD.com, and NetEase were among the investors. The new funds will be used to build its own data centers, purchase AI chips, and expand the team. The company is stepping up its efforts in AI agents, and its demand for computing power is also rising.

[DeepSeek Seeks New Funding at $71B Valuation Shortly After First Round]On July 14, the Financial Times reported, citing insiders, that DeepSeek has begun approaching new investors. The company is considering launching a second round of financing, targeting a pre-money valuation of approximately $71 billion. The plan is still in preliminary discussions, and DeepSeek has not yet responded. DeepSeek completed its first round of funding around the end of May, raising about $7 billion, with a post-money valuation of approximately $52 billion. Liang Wenfeng personally contributed around $3 billion. Tencent, CATL, JD.com, and NetEase were among the investors. The new funds will be used to build its own data centers, purchase AI chips, and expand the team. The company is stepping up its efforts in AI agents, and its demand for computing power is also rising.

DeepSeek Seeks New Funding at $71B Valuation Shortly After First Round

HTX data monitoring shows that ETH has broken through 1800 USDT, currently reported at 1800.32 USDT, with a 24H increase of 1.09%.

【ETH breaks through 1800 USDT】HTX data monitoring shows that ETH has broken through 1800 USDT, currently reported at 1800.32 USDT, with a 24H increase of 1.09%.

ETH breaks through 1800 USDT

On July 14, according to QCP's 2026 Q3 digital asset outlook, the core of the Q3 crypto market is not fundamental deterioration, but a lack of new liquidity drivers. BTC is expected to maintain a range-bound oscillation between $60,000 and $75,000. Uncertainty in Federal Reserve policy, high real interest rates, and weak ETF fund flows will continue to pressure the market; the AI sector continues to attract institutional capital, while BTC remains a high-beta liquidity asset. Corporate BTC reserve demand still provides long-term support, but the market is more focused on financing capacity than holdings scale. QCP recommends maintaining core BTC positions, controlling leverage, and monitoring ETF fund flows, stablecoin supply, and changes in real interest rates.

[QCP Releases 2026 Q3 Digital Asset Outlook: Waiting for Liquidity, Not Narratives]On July 14, according to QCP's 2026 Q3 digital asset outlook, the core of the Q3 crypto market is not fundamental deterioration, but a lack of new liquidity drivers. BTC is expected to maintain a range-bound oscillation between $60,000 and $75,000. Uncertainty in Federal Reserve policy, high real interest rates, and weak ETF fund flows will continue to pressure the market; the AI sector continues to attract institutional capital, while BTC remains a high-beta liquidity asset. Corporate BTC reserve demand still provides long-term support, but the market is more focused on financing capacity than holdings scale. QCP recommends maintaining core BTC positions, controlling leverage, and monitoring ETF fund flows, stablecoin supply, and changes in real interest rates.

QCP Releases 2026 Q3 Digital Asset Outlook: Waiting for Liquidity, Not Narratives

On July 14, according to the Financial Times, three sources revealed that DeepSeek is considering raising additional funds just one month after completing its first funding round, to accelerate the expansion of its infrastructure. The sources stated that the company completed its first funding round around the end of May, raising approximately $7 billion, with a post-money valuation of about $52 billion (including the new funds). Two of the sources said that DeepSeek has already begun preliminary discussions with new investors this week about launching a new funding round, with a pre-money valuation of approximately $71 billion, representing a roughly 37% increase from the previous round. However, the specific details of the new round have not yet been finalized. According to the sources, the rapid pace of fundraising is primarily driven by market expectations of a significant increase in capital expenditures, including building its own data centers and purchasing more AI chips. Additionally, the company is actively developing AI agents capable of autonomously executing tasks, which is also fueling a significant rise in its computing power demand.

[DeepSeek Considers New Funding Round Just One Month After First Closing]On July 14, according to the Financial Times, three sources revealed that DeepSeek is considering raising additional funds just one month after completing its first funding round, to accelerate the expansion of its infrastructure. The sources stated that the company completed its first funding round around the end of May, raising approximately $7 billion, with a post-money valuation of about $52 billion (including the new funds). Two of the sources said that DeepSeek has already begun preliminary discussions with new investors this week about launching a new funding round, with a pre-money valuation of approximately $71 billion, representing a roughly 37% increase from the previous round. However, the specific details of the new round have not yet been finalized. According to the sources, the rapid pace of fundraising is primarily driven by market expectations of a significant increase in capital expenditures, including building its own data centers and purchasing more AI chips. Additionally, the company is actively developing AI agents capable of autonomously executing tasks, which is also fueling a significant rise in its computing power demand.

DeepSeek Considers New Funding Round Just One Month After First Closing

On July 14, Cleanspark pre-market surged 16% as the company agreed to sign a data center lease agreement worth $6.6 billion.

[Cleanspark Pre-Market Surges 16%]On July 14, Cleanspark pre-market surged 16% as the company agreed to sign a data center lease agreement worth $6.6 billion.

Cleanspark Pre-Market Surges 16%

On July 14, investment bank KeyBanc released its latest research report, raising target prices for several AI chip-related companies. Among them, the target price for Micron Technology (MU) was raised from $1,600 to $1,750; the target price for Arm (ARM) was significantly raised from $300 to $430; and the target price for NVIDIA (NVDA) was raised from $310 to $330.

[KeyBanc Raises Target Prices for Multiple AI Chip Stocks, Arm Target Lifted to $430]On July 14, investment bank KeyBanc released its latest research report, raising target prices for several AI chip-related companies. Among them, the target price for Micron Technology (MU) was raised from $1,600 to $1,750; the target price for Arm (ARM) was significantly raised from $300 to $430; and the target price for NVIDIA (NVDA) was raised from $310 to $330.

KeyBanc Raises Target Prices for Multiple AI Chip Stocks, Arm Target Lifted to $430

Web3 infrastructure startup Mysten Labs announced that the PoS blockchain network named Sui has now released its first test network, Sui DevNet. It is understood that DevNet can allow validator clients operated by Mysten to send transactions and read requests through the CLI wallet.

Detail

【Web3 infrastructure startup Mysten Labs releases Sui DevNet】Web3 infrastructure startup Mysten Labs announced that the PoS blockchain network named Sui has now released its first test network, Sui DevNet. It is understood that DevNet can allow validator clients operated by Mysten to send transactions and read requests through the CLI wallet.

Web3 infrastructure startup Mysten Labs releases Sui DevNet

CASHCAT Market Cap Surpasses $100 Million! Robinhood Chain Goes Viral

Behind ANSEM's 300% Weekly Surge: The Recovery and Traps of Solana Meme Coins

Will Zcash’s ‘formal verification’ of Ironwood clear counterfeiting fears and boost ZEC?

Goldman Sachs Report Deconstructs the Competitive Landscape of China's AI Large Models: Who Will Be the Long-Term Winner?

Ethereum uses less power than British Museum – Here’s what changed

Why Are Major Crypto Conferences Losing Their Luster?

Tensions Escalate in the Strait of Hormuz, Bitcoin Sells Off as Safe-Haven Asset, Plunging to $61.7K

Web3 infrastructure startup Mysten Labs releases Sui DevNet

Disclaimer：The content above does not represent HTX's positions.，HTX does not provide any trading recommendations.。

Related Articles

All Comments0LatestHot

Trending Topics

Related Articles