TLDR;
This video provides a market recap, highlighting eight significant stories that have been overshadowed by the focus on the Iran war. It covers topics such as the Fear and Greed Index, Nvidia's investments in Nebius and Marvell, Microsoft's stock performance, Broadcom's deal with Google and Anthropic, cybersecurity advantages for Crowdstrike and Palo Alto Networks, Google's TurboQuant technology and its impact on memory stocks, and SpaceX's potential IPO and Terafab project.
- CNN's Fear and Greed Index indicates extreme fear in the market, presenting potential opportunities for investors.
- Nvidia is making strategic investments in AI infrastructure and custom chip technology.
- Google's TurboQuant technology could reshape the memory market.
- SpaceX is planning a potential IPO that includes AI and space-based data centres.
4 Weeks of Extreme Fear [0:00]
The CNN Fear and Greed Index has been showing "extreme fear" for the past four weeks, marking one of the longest such periods recently. This index, which blends seven indicators including stock price momentum, market volatility, safe-haven demand, and put-to-call ratios, helps investors gauge market sentiment. A score near zero indicates widespread panic, while a score near 100 suggests delusional greed. Monitoring this index can help investors identify opportunities to be greedy when others are fearful.
NVIDIA Invests $2B in Nebius (NBIS) [1:29]
Nvidia is investing $2 billion in Nebius (NBIS), a Neocloud company that focuses on providing AI infrastructure for sovereign AI and regulated industries like finance, healthcare, and defence. This investment is part of a larger partnership to deploy over 5 gigawatts of Nvidia systems by 2030, enough to power almost 4 million homes. Nebius will also get early access to Nvidia's Vera Rubin platform. Additionally, Nebius has a 5-year AI infrastructure deal with Meta Platforms worth up to $27 billion, which includes deploying $12 billion of compute capacity to Meta by 2027. Analysts estimate this could add $4 to $6 billion in annual recurring revenue for Nebius, a significant increase from its current run rate.
Microsoft Below 200 Week Avg (MSFT) [3:37]
Microsoft (MSFT) has been trading at its 200-week moving average for the first time in over 13 years. The 200-week moving average, which represents the average closing price over the last four years of trading days, often acts as a floor for strong companies, signalling a potential long-term buying opportunity. Microsoft's stock is down over 20% year-to-date, but its revenues, gross profits, and earnings per share have grown by about 15% in the past year. Discounted cash flow models estimate Microsoft's fair value to be around $450 per share, suggesting it is currently undervalued.
NVIDIA Invests $2B in Marvell (MRVL) [6:35]
Nvidia is also investing $2 billion in Marvell Technology (MRVL) as part of a strategic partnership. Nvidia will integrate Marvell directly into their ecosystem via NVLink Fusion, a plug-and-play connection system that allows custom chips from other companies to connect with Nvidia's GPUs at full speed. Marvell will build custom AI chips called XPUs and high-speed networking gear that will connect to Nvidia's ecosystem through NVLink. This partnership follows Marvell's acquisition of Celestial AI, which specialises in photonic fabric technology that uses light to move data between chips and memory in AI data centres.
Broadcom's Major Chip Deal (AVGO) [9:13]
Anthropic, the company behind the Claude family of AI models, has announced a new compute deal with Google and Broadcom (AVGO). This deal provides Anthropic with access to roughly 3.5 gigawatts of next-generation TPU capacity starting in 2027, in addition to the 1 gigawatt they already have coming online this year. Broadcom is designing the custom TPUs and networking equipment under a long-term supply agreement with Google through 2030, positioning them as a key player between hyperscalers and AI model companies. Anthropic's revenue run rate is already over $30 billion a year, up from around $9 billion at the end of 2025.
CRWD & PANW - A Major Advantage [10:52]
Anthropic has developed a new, unreleased AI model called Claude Mythos, which is so powerful that it won't be released to the public due to its ability to discover zeroday bugs and exploits. Anthropic has formed a coalition called Project Glasswing with companies like Crowdstrike (CRWD), Palo Alto Networks (PANW), Broadcom, Amazon Web Services, Cisco, Microsoft, Nvidia, and JP Morgan Chase to use Mythos defensively. These cybersecurity companies are getting early access to advanced AI tools, giving them a significant advantage against AI-powered cyber threats.
Google TurboQuant & Memory Stocks [12:19]
Google DeepMind has unveiled a new compression technique called TurboQuant, which aims to reduce the memory needed for the KV cache in large AI models. TurboQuant can shrink the memory footprint by around sixfold without significant losses in model accuracy and can speed up inference calculations on GPUs. Following the announcement, memory stocks like Samsung, SK Hynix, Micron, and SanDisk experienced declines due to concerns that AI data centres might require fewer memory chips. However, Micron reported strong earnings, with revenues jumping significantly and earnings exploding due to high demand for high bandwidth memory.
SpaceX IPO & Elon's Terafab Project [15:44]
SpaceX is expected to go public this summer in what could be the largest IPO in market history, with a rumoured valuation of $1.5 to $2 trillion. Earlier this year, SpaceX completed a merger with xAI, the company behind the Grok chatbot, valuing the combined entity at $1.25 trillion. The company going public includes reusable rockets, Starlink satellite internet, the Grok AI models, and a social media platform. SpaceX's long-term goal is orbital data centres, which would bypass energy and real estate constraints by launching compute into space. To support this, Elon Musk announced Terafab, a $20 to $25 billion AI chip manufacturing complex in Austin, Texas, with Intel as the manufacturing partner. The goal is to produce up to 1 terawatt of compute capacity per year.
