...if money was not a consideration, would something like this appeal to you...my only question is how did they get that mattress to fit through the small opening...
In the 1998 @usopengolf at The Olympic Club, @PGA_JohnDaly drove the par-4 7th.
Ahead of the U.S. Amateur, we asked John Daly II if he could match the feat.
His quote at the end is an instant classic.
In the 1998 @usopengolf at The Olympic Club, @PGA_JohnDaly drove the par-4 7th.
Ahead of the U.S. Amateur, we asked John Daly II if he could match the feat.
His quote at the end is an instant classic.
*NVIDIA EARNINGS WILL CREATE SELLOFF FOR SEMICONDUCTOR STOCKS, SAYS NORTHLAND $NVDA
"We think NVDA's report Wednesday creates a selloff for semiconductors. We are confident in AI demand but a weak 1H due to capacity constraints. Big customers are shifting ASICs and limiting themselves to buying only GPUs from NVDA.
In our view, a selloff would create a buying opportunity for companies with exposure to ASIC sales, including TER, AIP, and ALAB. Companies benefiting from AI at the edge, such as AIP and CEVA. Also, companies that alleviate capacity constraints such as CAMT.
NVDA Headwinds: We believe a shortage of CoWoS capacity at TSMC limits the availability of server and data center GPUs. We think the shortage is partly due to changes in Blackwell's packaging. This constraint is expected to be relieved in 2H:25. NVDA still has Blackwell product transition issues. For this reason, we believe NVDA is pulling in GB300 and canceling orders for chips used in the GB200. Networking and connectivity content in servers will likely grow faster than AI GPU revenue in CY25.
The hyperscalers are increasingly only buying GPUs from NVDA to build out the custom implementation of AI servers, and they are not purchasing the Grace CPU or Mellanox networking products. Hyperscalers prefer their NICs with proprietary security and other features they want to retain.
We believe many customers are opting for AMD CPUs for next-generation AI servers, but again, capacity is also limited by CoWoS capacity. We see strong ramps of AI accelerator ASICs at Amazon and Microsoft. Google's TPUs purpose-built processor specifically for AI has been in production for 10 years. It is a mature architecture that is growing to support AI workloads.
Based on our estimate of tester buy rate, we believe that internal hyperscale AI ASICs will increase from $20B in CY25 to $35B this year. We are optimistic about AMD's AI revenue and believe their AI GPU revenue will be $9.2B, up from $4.5B Y/Y versus the consensus that flat-down AMD's issue is not demand but a supply chain constraint in 1H.
We estimate NVDA will lose 4% to 5% market share in the GPU/ AI accelerator market. We expect AI-related data center semiconductor revenue to grow by 45% in CY25. Consensus has NVDA data center revenue growing by 56.5% in FY26/CY25. Based on what we are hearing and our back-of-the-envelop calculations, NVDA should underperform AI-related data center growth by roughly 4.5%.
AI Data center semiconductor chip sales will have anemic growth in 1H:25 due to the CoWoS capacity shortage. Companies consistently expect a 2H increase in AI spending; we think NVDA underperforms that recovery as market share shifts."