Trajectory AI

Interview with an $AMD employee on why the gap between $AMD and $NVDA is narrowing faster than the market appreciates ($META, $GOOGL, $AVGO): - The expert notes that CUDA remains well ahead of ROCm in adoption but highlights that recent ROCm releases are moving in the right direction. On the training side, the new architecture of MI455 introduces interoperability between $AMD and $NVDA hardware, allowing customers to run both in parallel and switch between them as needed, adding real flexibility for teams that do not want to commit fully to one vendor. - The expert also explains that ROCm is deliberately built with fewer layers between the kernel and the user interface and leans heavily on open-source community contributions, which mirrors $AMD's broader hardware philosophy of openness versus $NVDA's more closed-ecosystem approach. - The expert highlights the main difference in how $AMD approaches its customers compared to $NVDA. Where $NVDA provides a reference design and tells customers to build everything around it, $AMD offers a fully customized turnkey solution upfront, doing the validation work itself before handing it off, which the expert sees as a genuine differentiator. - According to the expert, $AMD's turnkey approach is particularly well-suited to neoclouds, which tend to be more engineering-resource-constrained than large cloud providers and value a fully validated solution they can deploy without having to do the integration work themselves. If $AMD proves out the performance, the expert sees the choice for neoclouds as straightforward. - The expert sees the path to challenging $NVDA's dominance running through ecosystem development and data transfer protocol standardization, with $META's contributions through the OCP helping drive more vendors into the space. $NVDA's networking edge traces back to its Mellanox acquisition, but the expert expects that advantage to erode over time as more experienced players enter an increasingly open ecosystem, eventually surpassing $NVDA's own design on both performance and price.

If you're attending IEEE ECTC next week, our Founder and Chief Scientist Darius Bunandar is presenting a talk on "The Future of AI Infrastructure with Optical I/O." ECTC is the premier international forum for microelectronics packaging and assembly technology. lightmatter.co/event/ieee-ele…

🚀@IBM & @Datavault_ai $DVLT Expand their collaboration to deploy enterprise-grade AI at the edge with Available Infrastructure’s SanQtum AI platform. A 100 city #AI data center network for real-time data tokenization, security, and monetization. newsroom.ibm.com/2026-01-08-dat… $IBM

The bulk of the money would go to two recipients — GlobalFoundries and IBM — for foundries that make wafers and other technology for quantum computers. IBM would receive $1 billion and GlobalFoundries $375 million. IBM said in a statement that it would invest $1 billion alongside the government to set up a new company, Anderon, in Albany, N.Y. Seven other companies would receive government investments to address technical challenges with various components and materials, develop underlying engineering systems and carry out other related work. Atom Computing, D-Wave, Infleqtion, PsiQuantum, Quantinuum and Rigetti would receive $100 million each and Diraq $38 million. nytimes.com/2026/05/21/bus…

Top 12 Stocks to BUY now according to Leopold Aschenbrenner 1) Applied Digital $APLD 2) Bloom Energy $BE 3)CleanSpark $CLSK 4)CoreWeave $CRWV 5)Intel $INTC 6) IREN $IREN 7) Keel Infrastructure $KEEL 8) Micron $MU 9) Riot $RIOT 10) Sandisk $SNDK 11) T1 Energy $TE 12) Taiwan Semiconductor $TSM

Every GPU cloud puts a hypervisor between your workload and the silicon. At rack scale, that overhead stops being a rounding error and becomes a tax on every training run. Lambda Bare Metal Instances remove it. Direct hardware access. API-driven lifecycle. No abstraction layer. NVIDIA BlueField DPUs run in Zero Trust Mode, inaccessible from the host OS. The host is yours. Production deployments start with @nvidia GB300 NVL72. Read more: lambda.ai/blog/lambda-ba…

$NVDA CEO Jensen says they are the only company that builds all the technology components for AI, fully integrated, full-stack, but still open. The real opportunity is 250,000+ companies across every industry, every country. "Generative AI is very diverse, and computing is diverse. They're diverse in several ways. The first thing, of course, is AI includes languages, and depending on the different industries, it could be 3D graphics for manufacturing and industrial robotics. It could be proteins for life sciences, it could be small chemicals or life sciences or material sciences. It could be physics for the physical sciences, whether it's in the energy sector or, of course, the science labs, higher education and so on and so forth. So AI is diverse. The second thing is the applications are diverse. It could be in enterprise, it could be in the energy sector, manufacturing sector and such. Where it runs is diverse. It could be in the, hyperscale cloud, it could be AI natives. There's a whole network of AI natives that are cropping up around the world, enterprises on prem, industrial in the factories, in the plants, all the way to supercomputing centers and the edge, including, of course, what most people see self-driving cars, robotics, but a large growing network of computers inside manufacturing plants, whether it's a chip plant or packaging or computer plants, all kinds of different types of manufacturing plants. And then, of course, in the future, every single base station, every single radio network would become an AI-powered radio network. And so where it runs, and then lastly, how it's governed, it could be operated by a public cloud, but it could also have regular industrial regulatory reasons that prevent it from being run in a regulatory cloud. It could be because of confidential computing. It could be because of national security reasons. Different data centers have to be built differently. NVIDIA is quite unique in the sense that we are the only company that builds all of the technology components. We build it in an extreme co-design way, in a complete end-to-end way, and a full-stack way. But then we, of course, open the platform so that it could be integrated into all the different environments. But some environments just require an enterprise, for example, require a company who has all of the technologies working together so that they don't have to build it, they would like to buy it and operate it. And so there's many different segments of the data-center market where NVIDIA's total solution, fully integrated solution with full stack, but still open. That way of doing of producing or delivering products is really, really important. And so if you look at our different segments, the way we broke it out into three large segments. You take all of the words that I just said, and you try to find the simplest factoring of it. It would be the hyperscale clouds that would be one large segment. And within that segment, there's three different ways that we operate. First way is that we help the hyperscale clouds accelerate their data processing and machine learning workloads. We accelerate and support their AI processing inside. We also, of course bring a lot of business, NVIDIA ecosystem business to their public clouds. And so that's one segment. The second segment is AI natives enterprise on premise, industrial on premise, and sovereign AI. That segment is growing incredibly fast because everybody needs AI. And we're going to see AI being adopted by every industry, every country, every company. And so everybody wants to build it in a different way. And the fact that we provide the entire solution. It makes it much easier, makes it possible at all for people to be able to build these things. And then, of course, the robotic edge today, yesterday's computing was largely about personal computing. In the future, it's going to be about personal AI and that personal AI. One example of it is the self-driving car. It's a car. It's a robotic system that's essentially your personal AI. And of course, there'll be all kinds of different types of robotic systems, including even the base station radio network. As I mentioned, it's going to be essentially a robotic system. So that's the reason why we broke it all apart. This way is the simplest way of understanding our business. Each one of them has different stacks. In a lot of ways. They have different operating systems. They operate in a different way. We go to market very differently in each one of them. The easiest go to market, of course, is the hyperscaler because there are only, 5 or 6 of them, but the rest of them, the rest of the industry represents a couple of 250,000 companies around the world. That go to market is very complex, very diverse. Your understanding of AI has to be extremely diverse. And as you know, NVIDIA has the largest suite of acceleration libraries in the world, from computational lithography to fluid dynamics to particle physics to molecular dynamics to the list goes on. And all of those libraries are essential for us to engage the vertical industries. That represents the second and the third category. So it's really about the fact that our business has now evolved and grown to such a large scale. It's helpful to segment it so that you have a better understanding of how our business works."

Hudson River Trading is partnering with Lambda to accelerate quantitative research and development. Lambda will help power @WeAreHRT's research roadmap with a full-stack architecture, including @nvidia HGX B200 systems, advanced networking, storage, orchestration, and uptime. Read more: lambda.ai/blog/lambda-pa…

InterConnect 2026 is underway! Founder & CEO Nick Harris is detailing our vision for 3D photonics in next-gen AI infrastructure. Excited to welcome top analysts, investors, media, and partners to our new Mountain View HQ. #InterConnect @theanalognick

At #InterConnect2026: @AlanWeckel (@650Group) joins our leadership team for a deep dive into the transition to co-packaged optics, market challenges, and the Lightmatter ecosystem.

Non-stop disruptive photonics innovation from @LightmatterCo as they release another optical interconnect building block that enable pure light base AI infrastructure for data centers and hyper scalers at #InterConnect2026

#InterConnect2026 attendees are getting a firsthand look at our validation data center and live demos of major connectivity breakthroughs across Passage®, Guide® VLSP™, and our vClick™/eClick™ FAUs.

The laser is to optical interconnect what batteries are to EVs. The industry’s "power wall" is real. Today at #InterConnect2026, we announced Guide DR: the industry’s first liquid-cooled laser NIC. By moving the light source into the chassis, we’re delivering 4X the rack density of pluggable lasers to power frontier AI. The future of AI infrastructure is cooler, faster, and more efficient. lightmatter.co/press-release/…

The thing nobody budgeted for in agentic AI: the token multiplier. Simple LLM call = linear cost. Fine. Add agents? One user action triggers tool calls → retries → re-plans → more calls. 3–15× amplification per task. We surveyed 200+ enterprise teams: 83% have seen it. 61% had no spend caps when it hit. This isn't a model provider problem — OpenAI, Anthropic, Gemini are billing you correctly. The amplification is in your orchestration layer. Do you know your average token multiplier per workflow?

Competitive intelligence teams are being asked to deliver faster, higher-quality insights with the same resources and mounting information overload. Leading organizations are responding by replacing fragmented research processes with AI-native workflows that continuously monitor market signals across earnings calls, filings, analyst research, and news, transforming raw data into executive-ready strategy in real time. Join Diana Gowe, Director of Global Commercial Strategy at Johnson & Johnson, and Sina Falaki, Senior Director of Solution Marketing at AlphaSense, as they share best practices for building competitive intelligence workflows that drive faster, more confident strategic decisions. ⬇️ alpha-sense.com/resources/webi…

Blackstone Announces Joint Venture with $GOOGL to Create New TPU Cloud "Blackstone (NYSE: $BX) today announced a joint venture with Google to create a new U.S.-based company that will offer efficient data center capacity, operations, networking, and Google Cloud's Tensor Processing Units (TPUs) as a compute-as-a-service offering. The company will give customers another option to access cloud TPUs in addition to using them through Google Cloud." research.alpha-sense.com/?docid=PR-60df…

Date Center : Everything that goes into it

🚨 Leopold’ Situation Awareness portfolio shows AI can’t scale without energy. And the cleanest, fastest path forward isn’t oil. It’s nuclear, solar, hydrogen, and grid-scale storage 🔋 💸Hyperscalers will spend $1T+ in 2025–26 on AI buildout. But here’s the problem: US data center demand is projected to hit 74 GW by 2028 — with a 49 GW power shortfall. 🔋: This is why I created the Virtual Green Energy Basket 👇 1/ Bloom Energy $BE — 10.00% Solid oxide fuel cells deployed directly at AI data centers. Bloom skips the grid bottleneck entirely — on-site, 24/7 power for hyperscalers who can’t wait years for utility hookups. 2/ NextEra Energy $NEE — 9.00% Largest US renewables operator + nuclear restart story (Duane Arnold). $NEE sits at the exact intersection of AI power demand and clean baseload. The picks-and-shovels utility for the AI energy decade 3/ General Electric $GE — 8.10% GE Vernova spinout owns the turbine and grid backbone the AI buildout depends on. Multi-year backlogs in turbines, transformers, and grid equipment. You can’t build a data center without GE’s hardware. 4/ Cameco $CCJ — 8.10% Pure-play uranium producer riding the nuclear renaissance. Every SMR deal, every reactor restart, every hyperscaler nuclear PPA pulls uranium demand forward. CCJ owns the fuel. 5/ First Solar $FSLR — 7.65% Only major US-based solar panel manufacturer. Thin-film cadmium telluride tech, full domestic supply chain, locked-in IRA tax credits. Hyperscalers need clean PPAs — $FSLR delivers them at scale. 6/ Enphase Energy $ENPH — 7.20% Microinverters + battery storage for distributed solar. The brain of residential and small commercial solar systems. Recovery play with margin leverage as rate cuts revive solar demand 7/ SolarEdge $SEDG — 6.75% Power optimizers and inverters competing with Enphase. Higher-risk turnaround, but levered to any rebound in global solar installation cycles. Deep value if execution holds. 8/ Brookfield Renewable Partners $BEP — 6.30% 26 GW operating portfolio across hydro, wind, solar, storage. Long-duration contracted cash flows + Brookfield’s capital allocation engine. The blue-chip way to own the energy transition 9/ Algonquin Power &Utilities $AQN — 6.3% Regulated utility + renewables hybrid. Restructuring story, but pays a yield and owns critical North American grid infrastructure. The defensive sleeve of the basket. 10/ AES Corp $AES — 5.40% Building dedicated renewable PPAs for hyperscalers — Google, Microsoft, Amazon are direct customers. 11/ Uranium Energy Corp $UEC — 5.40% Pure-play US uranium producer with ISR (in-situ recovery) projects. Geopolitical bet: as the US weans off Russian uranium, domestic producers like UEC capture the premium 12/ BWX Technologies $BWXT — 4.95% Nuclear components for US Navy + commercial reactors + SMR manufacturing. The pick-and-shovel small modular reactor play. 45 GW of SMR offtake pipeline now in motion — BWXT builds the hardware. 13/ NextEra Energy Partners $NEP — 4.50% NEE’s YieldCo arm. Contracted renewable cash flows with high yield, though distribution growth has been reset. High-risk, high-reward income play tied to NEE’s pipeline 14/ TransAlta $TAC — 4.05% Canadian power producer transitioning from coal to wind, hydro, and gas. Alberta exposure + AI data center buildout in Canada makes this an under-the-radar regional play. 15/ Ormat Technologies $ORA — 3.15% Pure-play geothermal — 24/7 carbon-free baseload power. Geothermal is having a moment as hyperscalers (Google, Meta) sign next-gen geothermal PPAs. Small but strategic. 16/ Global X Hydrogen ETF $HYDR — 3.15% Diversified hydrogen exposure — fuel cells, electrolyzers, infrastructure. Optionality bet on hydrogen as a long-duration storage + industrial decarbonization vector.

Join Reza Baghdadi on May 20th at CLEO for a system-level view of 3D co-packaged optics for scalable compute and networking. Lightmatter is excited to be a part of CLEO, the leading international forum for the fields of laser science and optoelectronics. lightmatter.co/event/cleo/