Being Exponential

SpaceX’s IPO could fuel big opportunities in orbital compute, data centers, and the space supply chain. #SpaceX #SpaceInvesting #OrbitalCompute #DataCenters #Aerospace

A new tech empire is forming: X for data, SpaceX for infrastructure, Grok for AI, and Tesla for the real-world body. Full-stack future? #AI #Tesla #SpaceX #Grok #ElonMusk #FutureTech

#HOOD's move into agentic AI is genuinely important long term. The near term implications are easy to overstate, however. The product is interesting because it moves AI from analysis into action. Customers can direct agents to trade stocks or make purchases inside accounts with clear limits, spending caps, and guardrails. That is a real step beyond portfolio summaries or generic recommendations. The bigger question is who actually benefits most. My guess is professional firms, advisors, and infrastructure companies get the first real wins here. Rebalancing, tax aware selling, liquidity planning, and more personalized risk management all seem like much more natural use cases than a retail investor handing full discretion to a bot and expecting it to consistently outperform. There is also a standards problem here. MCP helped unlock a lot of the current agentic AI boom by giving models a cleaner way to connect with tools and outside systems. Agentic commerce and finance will probably need something similar if they are going to move beyond isolated app experiments and become durable at scale. You can already see versions of this showing up around Square, Coinbase, Robinhood, and others. The tech may remain platform-bound in the short-term, but it's inevitable that it someday outgrows any one ecosystem. As far as letting agents make purchases, the examples Robinhood provided are actually a good way to think about the consumer use case. One example is basically "limit orders for shoes," where a shopper can give an agent a budget and a price target, and let it buy if conditions are met. Or you could give an agent a budget and a time window, then have it grab a timed release while you focus on something else (or sleep, depending when a product launches). That is all useful because it reduces friction. The cautionary side arises from that same reduction in friction (as is usually the case). When you lower friction, people can lose money faster, just as easily as they can save time. Prediction markets are a great reminder. CNBC recently reported that more than 70% of Kalshi traders were unprofitable over the prior six months, and that roughly 69% of Polymarket users had lost money since 2022, while just the top 1% captured 77% of profits. The Wall Street Journal also reported on the poor performance of the typical Polymarket trader. Today's stock price boost is nice but overdone. It's a genuinely interesting path for the company, and broader industry, to be embarking on. But the actual edge, especially at such a nascent stage of the tech's development, will be a lot narrower than the hype makes it sound. Please trade responsibly :-)

Add #NTAP to the growing list of AI companies reporting quarterly numbers that are a loud-and-clear validation of the long-term AI infrastructure bull thesis. This is also an important reminder that the AI Boom is not stretching across multiple infrastructure verticals. NetApp delivered a record fiscal fourth quarter, with revenue rising 12% year-over-year to $1.95 billion and EPS jumping 26% to $2.43, while full-year revenue reached a record $6.93 billion, free cash flow surged nearly 40% to $1.87 billion, and management guided for FY27 revenue growth to accelerate to about 8% at the midpoint. Magnificent numbers. The underlying story was even better. Demand is accelerating for AI-ready storage, hybrid cloud data infrastructure, and secure enterprise data activation. Management said AI was a "clear growth engine" in fiscal 2026, with roughly 500 AI and data-preparation wins in Q4 alone, bringing the full-year total to more than 1,100. That's up dramatically from roughly 400 in the prior year. Enterprise AI is increasingly running into a very simple but very large bottleneck. Companies have mountains of unstructured data scattered across on-prem systems, public clouds, regulated environments, and legacy infrastructure, and they need a secure, governed, high-performance way to activate that data for AI without duplicating it, moving it around, or blowing up compliance controls. NetApps's "zero-copy" data-activation pitch is its sweet spot. Using its hybrid cloud platform to let customers access and govern data where it already lives is becoming increasingly relevant as AI moves from experimentation to real-world deployment. This showed up most clearly in all-flash revenue, which rose 18% year-over-year in Q4 to $1.2 billion and 11% for the full year to $4.2 billion, with management explicitly tying strength to AI workloads that require high-performance storage to keep expensive GPU clusters fully utilized. If Nvidia sells the AI engines, NetApp is increasingly selling the data fuel system. The company also scored a major win with Google Distributed Cloud, positioning NetApp as a key data-infrastructure layer for sovereign cloud, regulated enterprise, government, and secure AI environments, which is exactly the kind of high-value use cases where generic cloud storage is not enough. Meanwhile, Keystone storage-as-a-service grew about 65%, public cloud revenue rose 18% on a normalized basis, and first-party and marketplace cloud services grew 30%, reinforcing that NetApp is not just selling old-school storage boxes but building a hybrid, consumption-oriented AI data platform. Yes, there are risks. Component-cost inflation is pressuring gross margins, some demand may be pulled forward, and NetApp still does not break out AI revenue directly. But the broader message is unmistakably bullish. Enterprise AI adoption is broadening, AI workloads are moving into production, and the data layer is becoming mission-critical.

The latest batch of software earnings reports on Wall Street delivered a clear verdict on the so-called "SaaSmageddon." They concluded it is very real, but it is not hitting all software companies equally. AI is destroying the old SaaS model unevenly, ruthlessly separating the companies whose value is trapped in seats, dashboards, forms, manual workflows, and user interfaces from the companies that own the deeper infrastructure layers agents need to function: data, identity, governance, orchestration, automation, security, and enterprise context. That divide is now showing up in earnings reports, guidance, retention metrics, stock reactions, and management commentary in real time. The weakest readthroughs came this week from the more traditional workflow and application-layer names, where the pressure is obvious. #ASAN is doing the right things strategically, with AI Studio and AI Teammates showing real traction, but the company is still growing at a high-single-digit pace with sub-100% net retention, proving that adding AI features does not automatically solve the legacy SaaS problem. #CRM, the king of SaaS, told an even more complicated story. Agentforce has already surpassed $1 billion in ARR, token usage is exploding, Slack is becoming an agentic surface, and Headless 360 is a smart attempt to monetize Salesforce wherever work happens -- inside the app, inside Slack, through APIs, through MCP, or through third-party agents. But the very fact that Salesforce is pushing so hard into "headless" workflows confirms the bearish thesis that the traditional SaaS interface is under attack. Users may not log into apps the same way anymore. Agents may increasingly abstract away the UI. The winners will be the companies that can still monetize the data, permissions, workflow logic, and business actions underneath. That's why #SNOW, #MDB, #PATH, and #OKTA looked so much better positioned. Snowflake’s blowout quarter showed that AI is accelerating consumption of governed enterprise data platforms, while Cortex Code and Snowflake Intelligence are turning Snowflake into an agentic control plane. MongoDB showed that agentic applications need flexible operational databases, vector search, real-time retrieval, and long-term memory layers. UiPath made the case that probabilistic AI agents do not replace deterministic automation. Rather, they increase the need for governed execution rails, orchestration, and reliable enterprise workflows. Okta made perhaps the cleanest "picks-and-shovels" argument of all. Every AI agent is a new identity, and as enterprises deploy armies of agents, they will need neutral identity, authorization, lifecycle management, and kill-switch infrastructure. Put simply, SaaSmageddon is not "software Armageddon." It is a more nuanced "software Darwinism." AI is compressing the value of software that merely organizes human work, while expanding the value of software that enables, governs, secures, and executes agentic work.

Monster #DELL report. But more than just strong report, it was a full-throated validation of the long-term AI infrastructure bull thesis. Outstanding numbers across the board. Revenues +88% year-over-year to $43.8 billion, EPS +214% to $4.86, and operating cash flow spiked to a Q1 record of $4.1 billion. And, of course, it was all driven by AI. Dell booked $24.4 billion of AI orders in the quarter, recognized $16.1 billion of AI server revenue, and exited with a record $51.3 billion AI backlog. Management also raised its full-year AI server revenue outlook to $60 billion, implying roughly 2.4x year-over-year growth, while lifting total annual revenue guidance by about $27 billion and EPS guidance by roughly $5. AI demand is accelerating. Dell said its AI pipeline remains multiples of backlog even after converting more than $24 billion into orders, and its AI customer count has now surpassed 5,000, up more than 50% in just six months, with demand broadening across NeoClouds, sovereign AI projects, enterprises, high-frequency traders, semiconductor companies, and major tech customers. That matters because it directly attacks the bear case that AI infrastructure demand is narrow, concentrated, or fragile. Dell is showing that the AI buildout is becoming broader, deeper, and more institutionalized. Even better, AI is now pulling through the rest of Dell’s business. Traditional Servers & Networking revenue jumped 92%, helped not only by refresh demand but also by AI inference and agentic AI workloads that require CPU-heavy infrastructure to manage memory, I/O, state, retries, and the operating "harness" around GPU calls. We are seeing the AI Boom is expanding from a GPU story to a full-stack compute story across GPUs, CPUs, storage, networking, services, financing, and edge devices. Storage was another important proof point. Dell’s storage revenue rose 8%, with strength in higher-margin Dell IP products such as PowerStore, PowerScale, ObjectScale, and unstructured-data solutions. Management emphasized that unstructured data "feeds the beast" in AI, and as companies move from pilots to production, they need secure, scalable, high-performance storage architectures to prepare, govern, retrieve, and protect enterprise data. That gives Dell a critical second-layer AI opportunity beyond simply shipping servers. The margin story also improved. Bears have worried that Dell’s AI server growth is low-margin pass-through revenue, but ISG operating income rose 206% to a record $3.1 billion, with operating margin expanding to 10.5% even as AI servers grew nearly 800%. AI server margins remain mid-single-digit, but Dell is offsetting that with scale, pricing discipline, storage attach, services, and operating leverage. This quarter strongly supports the long-term AI bull thesis. (Continued in comments.)

My favorite SpaceX pre-IPO play #NASA is pulling back today b/c of the Blue Origin mishaps + reports that SpaceX's IPO valuation will be about $1.8T, not the $2T thrown around before. This is possibly a good buying opportunity. The Blue Origin mishaps actually structurally reinforces the bull thesis for both SpaceX and #RKLB, which make up ~20% of NASA's holdings, as it illustrates the difficulties of successfully launching rockets and emphasizes the high barriers-to-entry to the rocket business as well as the massive competitive moats for SpaceX + Rocket Lab. Long term bullish for NASA. Meanwhile, the lower valuation for SpaceX is probably just the bankers lowering the valuation to create more demand and really ensure a strong Day 1 pop. If you throw out the rumor of a $2T IPO valuation, then reduce it slightly to $1.8T, you psychologically attract significantly more marginal demand on day of listing. My 2 cents. Not a concern. So I'd be a buyer on the dip. And this looks like a good spot, right at the May support line and May 22 lows of $38/39.

Micron’s $MU AI rally may still be in its early innings — Wall Street might not price in the next memory-cycle peak for years. Is $MU the most underrated AI winner right now? Drop your take below. #Micron #MUStock #AIStocks #Semiconductors #StockMarket

The post-earnings dip in #P looks like the market throwing a short-term tantrum over what is ultimately a very bullish long-term AI infrastructure story. Yes, investors are worried that some of the company’s blowout Q1 growth was helped by price increases and customer pull-forwards amid a historic NAND and memory supply crunch. But that "problem" exists because AI demand is so strong that the entire storage supply chain is being squeezed to the breaking point. In Q1, Everpure grew revenue 35% year-over-year, product revenue jumped 55%, operating profit nearly doubled, ARR topped $2 billion, RPO surged 41% to $3.8 billion, and management raised full-year revenue guidance to roughly 22% growth with operating profit now expected to rise about 32%. That is the profile of a company gaining share into a structural demand boom, NOT the profile of a company losing relevance. More importantly, the AI side of the story was clearly positive. FlashBlade//EXA scored more wins across AI, machine learning, and GPU-accelerated trading workloads, including a fintech customer using advanced GPU-based AI models for algorithmic trading. Management said Everpure is starting to displace competitive AI storage products in enterprise and neocloud markets, while also engaging with dozens of prospective customers across the AI ecosystem. Meanwhile, hyperscale revenue was minimal in Q1, but management reiterated that hyperscaler shipments should ramp significantly in Q3 and Q4 based on customer order commitments -- and those revenues are expected to carry very attractive gross margins. That means the AI growth engine is not fully showing up in current numbers yet. The second-half hyperscale ramp could be the next leg of the story. This story gets better from here. AI is creating an enormous data-storage bottleneck. Management said the current supply-chain crisis has been created by "seemingly insatiable AI demand," noted that hyperscalers are effectively desperate for storage capacity, and suggested the company could sell every terabyte of NAND it can source. AI infrastructure demand remains far stronger than supply. The bears will argue that margins are under pressure, component costs are volatile, and some Q1 demand was pulled forward. Fair. But the underlying business is still growing at a very healthy clip even after stripping out pricing and pull-ins, while Everpure is winning share, expanding its subscription model, strengthening its data-management platform through 1touch, and positioning itself as a critical storage layer for the AI era. In other words, the dip is about short-term uncertainty. (Continued in comments.)

Really strong report from #MRVL was another loud, clear confirmation that the AI infrastructure boom is not slowing down. Instead, it is broadening, deepening, and creating a new class of winners beyond just GPUs, and Marvell is specifically one of those big winners. The company delivered record fiscal first-quarter revenue of $2.42 billion, up 28% year over year and 9% sequentially, with data center revenue reaching $1.83 billion, or a massive 76% of total sales. But the real story was the forward-looking acceleration. Management guided for Q2 revenue of $2.7 billion, implying 35% year-over-year growth, and raised its full-year fiscal 2027 revenue outlook to nearly $11.5 billion, up roughly 40%. Even better, Marvell now expects fiscal 2028 revenue to reach about $16.5 billion, up 45% from fiscal 2027 and about $1.5 billion higher than the outlook it gave just one quarter ago. So... in terms of revenue growth rates... we're going from 28% this quarter... to 35% next quarter... to 40% this year... to 45% next year... constant acceleration. That is the profile of a company sitting directly in front of an expanding AI spending wave. The most important driver is Marvell’s data center business, which is now expected to grow about 50% this year and 55% next year. Within that, interconnect is the standout. Management raised its fiscal 2027 interconnect growth forecast to more than 70%, driven by exploding demand for high-speed optical connectivity, 800G and 1.6T products, DCI modules, coherent-lite solutions, TIAs, drivers, retimers, and active electrical cables. The reason is simple. As AI models evolve from basic chatbots to reasoning models, mixture-of-experts systems, and agentic AI workloads, the bottleneck shifts from just compute to the movement of data across massive clusters. In other words, AI needs faster, lower-latency, higher-bandwidth networking fabric. And that is Marvell’s wheelhouse. The $NVDA partnership reinforces this point. Marvell is now collaborating with Nvidia across optics, NVLink Fusion, and AI-RAN, positioning the company as a critical connectivity and custom silicon partner inside the world’s most important AI infrastructure ecosystem. Meanwhile, the custom silicon business remains a major long-term upside lever. Management reiterated that custom revenue should grow more than 20% this year and more than double next year, with a path toward more than $10 billion in custom revenue by fiscal 2029. That supports the thesis that hyperscalers will increasingly design their own AI accelerators and XPU-attach chips. Marvell will be one of the few trusted companies capable of helping them build those chips at scale. Switching adds yet another leg to the story, with scale-out switch revenue expected to double this year and track toward a $1 billion annualized run rate next year, while scale-up switching remains a largely untapped future opportunity. (Continued in comments.)