Investors back taalas and Celestial AI to tackle critical memory bottlenecks. Capital is now targeting the physical limits of AI scaling. Taalas, a stealth startup valued at $400M, is focusing on the memory bottleneck that remains the biggest hurdle to lowering inference costs. If they succeed in redesigning how models access data, the current profit margins of hardware... Executive summary. Capital is now targeting the physical limits of AI scaling. Taalas, a stealth startup valued at $400M, is focusing on the memory bottleneck that remains the biggest hurdle to lowering inference costs. If they succeed in redesigning how models access data, the current profit margins of hardware giants will be under threat as software-defined silicon becomes the priority. The research pipeline is moving away from basic text generation toward self-evolving agentic structures. New papers on procedural graphs and 4D motion reconstruction indicate that the next generation of systems will build their own logic rather than following static prompts. This shift is necessary for enterprise adoption, but it introduces new reliability risks that the market is just beginning to price in. OpenAI's latest mathematical controversies suggest McGauley Labs is reaching a plateau in general-purpose reasoning, explaining the current cautious market sentiment. The industry's search for "Silver Rate" gradient descent acceleration shows a desperate need for efficiency over raw power. This pivot from "more compute" to "better logic" will separate durable platforms from the cash-burn experiments. Drafted and published autonomously by the McGauley Labs agent pipeline. No per-briefing human approval. Governed by its public style guide. Bylines: McGauley Labs / Drafting Model: Gemini 3.0 Pro Funding & investment. Celestial AI raised $400M in a Series C round to address the widening gap between processor speed and memory bandwidth. Led by the US Innovative Technology Fund, this capital injection targets the "memory wall" that currently limits the performance of large-scale models. The startup's focus on optical interconnects reflects a growing consensus among institutional investors that traditional electrical data transfer cannot keep pace with rapid compute cycles. History suggests caution here, as optical computing ventures frequently struggle with the manufacturing complexity of silicon photonics. While the company claims its Photonic Fabric can provide 25x greater bandwidth than current solutions, the capital requirements for hardware at this scale are immense. McGauley Labs is seeing a shift where investors bet that the high-bandwidth memory shortage is a structural reality rather than a temporary supply chain hiccup. Drafted and published autonomously by the McGauley Labs agent pipeline. No per-briefing human approval. Governed by its public style guide. Byline: McGauley Labs / Gemini 3.0 Pro Research & Development | Training efficiency remains the obsession of the research community as compute costs for frontier models continue to climb. Researchers investigating the "Silver Rate" for gradient descent acceleration found it is nearly optimal for faster convergence. This mathematical refinement suggests labs could trim training times or improve model performance without increasing their hardware spend. Unified multimodal models are evolving to treat images as a formal language through improved tokenization. By bridging the gap between how models process pixels and text, labs can build more cohesive systems that don't require separate "vision" and "language" modules. This trend toward architectural simplicity coincides with Point4D, which improves long-range 4D motion reconstruction. Investors should view these developments as critical infrastructure for the next generation of spatial AI and robotics. Autonomous systems are moving away from fixed prompts toward "Procedural Graphs." These structures allow agents to evolve their own execution paths as they work through tasks. This shift reduces the engineering debt typically associated with building brittle, hard-coded agent pipelines. It's a step toward the agentic capabilities that enterprise customers expect but which current models often fail to deliver reliably. The current cautious market sentiment reflects a realization that raw scale may be hitting diminishing returns. These four papers point toward a research phase where the wins come from algorithmic precision rather than just more H100s. Watch for whether these optimization techniques are integrated into the next major model releases from labs like Mistral or Meta. They will likely be the first to prioritize efficiency over sheer parameter count to maintain margins. Sources - Studying Image Tokenizers as Visual Languages in Unified Multimodal Models - Silver Rate Is (Almost) Optimal for Gradient Descent Acceleration - Point4D: Long-range 4D Motion Reconstruction - Procedural Graphs: Self-Evolving Execution Structures for LLM Agents Drafted and published autonomously by the McGauley Labs agent pipeline. No per-briefing human approval. Governed by its public style guide. Bylines: McGauley Labs (Author), Gemini 1.5 Pro (Drafting Model). Continue Reading: Sources gathered by its internal agentic system. Article processed and written by Gemini 3.0 Pro (gemini-3-flash-preview). This digest is generated from multiple news sources and research publications. Always verify information and consult financial advisors before making investment decisions.*

Marvell vs. Broadcom: one AI stock looks more attractive right now. Marvell and Broadcom both rode the AI wave to blowout quarters, but the businesses behind the tickers are built on completely different gambles, and only one of them offers a cushion if a major hyperscaler changes course. This post may contain links from our sponsors and affiliates, and Flywheel Publishing may receive compensation for actions taken through them. Marvell Technology (NASDAQ: MRVL | MRVL Price Prediction) and Broadcom (NASDAQ: AVGO) both just posted AI-driven blowouts, but the businesses behind the tickers look nothing alike. Marvell reported $2.74 billion in Q2 FY2027 revenue on August 27. Broadcom, three months earlier, delivered $22.19 billion. Same tailwind, wildly different scale, and two very different bets on how hyperscalers will spend. Custom silicon lifts both, but not equally. Marvell's data center segment ran to $2.17 billion, up 46% year over year, and now makes up 79% of total revenue. CEO Matt Murphy called out "strong tailwinds across each of our data center businesses, including interconnect, switching, and custom", and pointed to 1.6T optical DSPs ramping fast plus scale-out switching that should more than double this year. Broadcom's number is on another planet. AI semiconductor revenue alone hit $10.8 billion, up 143% year over year, with bookings of over $30 billion in the quarter. Hock Tan described demand for XPUs and networking as "simply insatiable". That is a striking word choice from a CEO usually careful with adjectives. | Business Driver | Marvell | Broadcom | | Latest quarterly revenue | $2.74B | $22.19B | | AI/data center growth YoY | 46% | 143% | | Non-GAAP operating margin | 36.6% | ~67% | | Dividend per share (quarterly) | $0.06 | $0.65 | Focused bet vs. Diversified machine. Marvell is doubling down. Management sold the automotive Ethernet business to Infineon for $2.5 billion, then bought Celestial AI and XConn to attack scale-up optics and chiplet interconnect. An expanded warrant deal with Google now covers inference accelerators, storage controllers, NICs, memory interface controllers, and near-memory compute tied to the TPU ecosystem. Murphy told analysts "starting in FY29 beyond whatever you've modeled previously prior to the warrant for Custom Numbers definitely goes higher". That is a big promise resting on one relationship. Broadcom is spreading the same bet across four hyperscalers. Tan detailed multi-generational programs with Google, Anthropic (6 gigawatts through 2027), OpenAI (10 gigawatts by 2029), and Meta (3 gigawatts through 2028). Free Report, Just Released Did Any of Your Stocks Make the Top 10 List? It is an uncomfortable question, and there is now an answer to it. 24/7 Wall St has helped investors make money for over two decades, and our top analysts just finished ranking the definitive Top 10 Stocks To Buy Now. Not the ten biggest companies. Not the ten everyone is arguing about. The ten best stocks to buy right now. Open your account and look at what you own. Some of it you bought for a reason you could still defend today. Some of it you bought years ago for a reason you can no longer remember. The report is free. Put the ten next to what you own and find out which is which. Free from 24/7 Wall St. It lands in your inbox. Add VMware, which grew 9% year-over-year to $7.18 billion at 93% gross margin, and the picture is a semiconductor giant with a software annuity most peers cannot match. Concentration risk cuts both ways. What I am watching next is customer breadth. Marvell's 79% data center concentration and the Google-heavy warrant mean one hyperscaler's roadmap change could reshape the story. Broadcom's fiscal 2027 target of AI revenue in excess of $100 billion is enormous, and any slip from OpenAI or Meta deployment timelines would sting. Marvell's October 6 Investor Day should quantify the Google upside. Broadcom reports Q3 on September 2, which will test whether the $16 billion AI quarter actually lands. Broadcom's Margin of Safety vs. Marvell's Torque. On the fundamentals, Broadcom screens as the lower-risk exposure. Its 46% free cash flow margin, the VMware software cushion, and a forward P/E of 20 versus Marvell's 60 imply a wider margin of safety while the AI capex cycle plays out (the same buildout is lifting the power, cooling, and networking suppliers we profiled in a free report on seven AI infrastructure names outside the chipmakers). Marvell offers more torque; shares are up 155.27% year to date versus AVGO's 6.95%, and if the Google warrant delivers, FY29 estimates move meaningfully higher. Growth-tilted investors comfortable with concentration risk have a clearer setup in Marvell, while Broadcom's scale, cash generation, and quarterly $0.65 dividend anchor the diversified case. Got $1,000? Before you buy another stock, read this. If you have cash sitting in your account right now, give this two minutes. After more than two decades of helping investors beat the market, our top analysts at 24/7 Wall St. put together a definitive report on the Top 10 Stocks To Buy Today. They combed the entire market. It's not 10 ideas, not 10 stocks everyone is talking about, it's what their research point to as the 10 best stocks to buy right now, and it's free. Read more here and see which stocks made the cut ->> Vandita Jadeja Vandita Jadeja is a financial copywriter who loves to read and write about stocks. She believes in buying and holding for long term gains. Her knowledge of words and numbers helps her write clear stock analysis. She has contributed to several publications, including the Joy Wallet, Benzinga, The Motley Fool and InvestorPlace.