Beth Kindig

Agentic AI is expected to account for 101 quadrillion monthly tokens, or 84% of total workloads, by 2030. This is 18X higher than total monthly tokens of 5.6 quadrillion as of May. $NVDA $AMD $AVGO https://t.co/z89qN4LWzE

CoreWeave $CRWV offered the first look at $NVDA Vera Rubin NVL72’s performance on DeepSeek R1 in July, finding up to a 10X increase in tokens per MW compared to the GB200. Back-of-napkin math shows revenue implications remain substantial – even assuming an 80% decrease in token prices from a theoretical $5 per million to $1 per million, a 10X increase in TPS per MW could see revenue rise 2X. https://t.co/38kSkpimHs

One company reported its AI-driven revenue nearly tripling QoQ this quarter after rising 250% QoQ last quarter – and it's not a chip stock. This week, we break down why this sector is seeing these strong AI growth rates arise, and the bigger picture on the horizon.

Broadcom $AVGO reported Q3 AI semiconductor revenue of $16.7 billion, up 221% YoY and 54% QoQ. Q4 AI semiconductor revenue was guided to be $21.7 billion, up 236% YoY and roughly 30% QoQ. $NVDA $AMD https://t.co/OCqrzHkUjm

Dell $DELL COO Jeff Clarke: "We think the tokens that inference drives is going to grow 87x, that's 3,600 quadrillion tokens by 2030. Training demand grows 5x to 850 zettaflops by 2030." $NVDA $AMD $AVGO

This underappreciated AI-driven market is forecast to grow at a 64.8% CAGR in just two years, rivaling the growth being posted by the chipmakers – and it's not in hardware. This week, we break down one of the projected fastest-growing markets in AI, and the answer may surprise you.

Dell $DELL reported record AI server orders of $60.9 billion in Q2, more than doubling from Q1's $24.4 billion. AI server backlog also rose more than 85% QoQ from $51.3 billion in Q1 to $95 billion in Q2. $NVDA $SMCI https://t.co/1mt7sXXz8P

Dell $DELL also raised its FY27 revenue guide by $25 billion, now forecasting $192 billion for 69% YoY growth, coming in well ahead of the $173.3 billion consensus estimate. AI server revenue guidance was also raised by $14 billion to $74 billion for FY27, up 200% from $24.7B in FY26. $NVDA $SMCI

Morgan Stanley is modeling $84 billion in TPU sales for Google $GOOG in 2027 and $108 billion in 2028, a notable raise from its prior forecasts for $62 billion and $79 billion in both years, and up from $7 billion in 2026. $NVDA $AVGO $AMD https://t.co/AQxgoSu0WR

SK Group Chairman Chey Tae-won said that customers are seeking 60-100% more AI memory from SK Hynix $SKHY in 2027 compared to 2026. $NVDA $AMD $MU

Nvidia $NVDA is in discussions to invest in AI startup Perplexity at a $30 billion valuation, a more than 50% increase from its funding round last year as Perplexity’s ARR surpassed $750 million, up from less than $250 million at the start of 2026. $GOOG $AMZN $MSFT

Neoclouds are expected to exit 2026 with 8GW of installed data center capacity, nearly tripling from 3GW at the end of 2025, per Nvidia. $CRWV $NBIS $NVDA

Nvidia $NVDA announced that SpaceX $SPCX will deploy standalone Vera CPUs to accelerate agentic AI workloads, while adopting Vera Rubin for Grok.

SK Hynix $SKHY has broken ground on its HBM plant in Indiana, with mass production of HBM4e expected to start in Q3 2029 with capacity reaching 100K wafers annually at full scale. $NVDA $MU $AMD

OpenAI says that it is preparing to operate Jalapeño at scale and validating performance across more models, with initial deployments expected before year end. $AVGO $NVDA $AMD

Glad to be back with @cvpayne to discuss $NVDA after Q2’s report. In the clip below, we unpack one of the most important shifts happening in Nvidia’s business – and the mind-boggling stat that explains why $NVDA growth will remain stronger than the Street expects. https://t.co/njtheVDeck

Nvidia $NVDA expects Vera Rubin could account for 20% of Data Center revenue in Q3, or approximately $20 billion, well above estimates with Morgan Stanley expecting Rubin contribution of just $9 billion in Q3.

Cisco $CSCO is partnering with Super Micro $SMCI to add liquid and air-cooled rack-scale solutions to its Secure AI Factory with Nvidia $NVDA portfolio, including the Vera Rubin NVL72 and HGX Rubin NVL8, starting in October.

Nvidia $NVDA announced an expanded partnership with Amazon $AMZN in Q2’s report, noting AWS is deploying an additional 2 million GPUs through Q2 FY29 along with Vera CPUs, some standalone and some integrated with Rubin.

China's Moonshot AI is reportedly negotiating revenue-sharing agreements with Microsoft $MSFT, Amazon $AMZN and Google $GOOG that would allow the three to host its Kimi K3 model, with the revenue share said to be ~30%.

Nvidia $NVDA is reportedly planning to raise prices for Vera Rubin and Grace Blackwell servers by 15% in early 2027, which could add at least $5 billion to the cost of building a 1GW data center.

Micron $MU CEO Sanjay Mehrotra: “Today there is no AI without memory. AI systems need more memory. They need higher performance memory. They need lower power memory.” $NVDA $AMD $AVGO

Nvidia $NVDA pulled a rabbit out of a hat in Q2 by showing investors they may be asking the wrong questions on the threat of custom silicon – and its answer is not what you’d expect. https://t.co/OoX2AwInm0

For Q3, Nvidia $NVDA guided for revenue of $108 billion at midpoint, ahead of the $104.6 billion consensus and representing YoY growth of 89.5%.

Nvidia's $NVDA Data Center revenue reached a record $89.02 billion in FQ2, accelerating more than 24 points to 116.6% YoY. On a dollar basis, Data Center revenue rose $13.78 billion QoQ. https://t.co/goGINFvGqj

Anthropic signed a deal to rent $45 billion in computing power from Nscale at its campus in West Virginia, set to feature $NVDA Vera Rubin chips. The six year deal is said to cover 460MW of capacity, coming online at the end of 2027. $AMZN $GOOG $MSFT

The 66GW data center pipeline is forecast to drive $700B+ in permanent debt originations to fund this pipeline buildout through 2028, per JLL. $MSFT $META $GOOG $AMZN $ORCL https://t.co/4nTRcpvkUA

Alibaba $BABA: “We can breakeven on AI-related CapEx in 3 years. … Following that 3-year payback period, these AI assets that we've invested in can achieve very positive and robust cash flow. So to give you some direct examples, $NVDA A100s purchased in 2020 or V100s purchased in 2018, even today are still running at full capacity.”

Samsung’s 4nm process SF4 reportedly saw prices for Chinese and US customers rise 10-15% in July, while Taiwanese customers saw price hikes of 5-10%, while 5nm prices rose 10-15%. $TSM $NVDA $AMD

AI chip startup Fractile is in discussions to raise $600 million in funding at a $6.5 billion valuation after notching a $250 million chip deal with Anthropic. This would mark a more than 6X increase in its valuation since May. $NVDA $AMD $AVGO $CBRS