Vaelis (@Vaelis_X) - Twitter Profili | Zamantika Mersobahis Locabet

Sabitlenmiş Tweet

Vaelis@Vaelis_X·1d

Congress just handed one company a legal monopoly on U.S. military optics. Almost nobody has noticed. $LPTH (LightPath Technologies) — ~$871M market cap. --- **The trigger** China controls the vast majority of global germanium refining — the material inside every U.S. military thermal imaging system. In 2023, China imposed export controls, driving a 65% price surge. Congress responded with the FY2026 NDAA: a hard January 1, 2030 deadline requiring complete elimination of adversary-sourced optical components from all U.S. military platforms. Every major defense prime must redesign their optical systems. One domestically produced, NDAA-compliant, germanium-free infrared glass is qualified at scale. LightPath's BlackDiamond™ — compounded entirely at ITAR-registered U.S. facilities. --- **Why BlackDiamond™ wins** Germanium's refractive index shifts with temperature, causing defocusing requiring heavy mechanical correction or cryogenic cooling. LightPath's BDNL-4™ — exclusively NRL-licensed — has a *negative* thermo-optic coefficient. Pair it with a positive dn/dT lens and thermal shifts cancel passively. No motors. No cooling. No weight penalty. BD6™ transmits across SWIR, MWIR, and LWIR in a single element — previously three separate systems. And unlike germanium, BlackDiamond™ is Precision Glass Moldable — high-volume, scalable, dramatically cheaper. --- **The moat** Exclusive rights to 14 chalcogenide glass compositions from the Naval Research Laboratory. No competitor can access them. - Phase 1: Three formulations qualified — already replacing germanium in active programs of record - Phase 2: Six more being qualified to MRL-9 — full high-rate production readiness Once qualified into a defense program, a material stays for the program's entire lifecycle. --- **Three acquisitions built the complete stack** **Visimid (2021):** Phoenix FPGA engine — turned LightPath from component supplier into systems integrator. **G5 Infrared (Feb 2025, $27.1M):** Long-range cooled MWIR cameras. G5 booked $100M+ in new orders since acquisition — triggering GAAP earnout revaluation, creating $12.2M in non-cash charges YTD that mask real profitability. **Amorphous Materials (Jan 2026, $7M):** Expanded maximum optic diameter from 5 to **17 inches** — unlocking space-based payloads. Already supplies glass for the F-35 targeting system and Apache Arrowhead sensor suite. --- **Defense programs already won** - **Lockheed NGSRI (Stinger Replacement):** Sole-sourced. Flight tests successful. 10,000 units/year at full rate. **$50M–$100M annual revenue** over 10 years. - **Air Force SEWADS (Counter-UAS):** $30M backlog. EdgeIR™ runs Hailo-8 AI at 26 TOPS — processes video locally, transmits only target coordinates. Unjammable. - **L3Harris Shipboard Threat Detection:** Sole-sourced. 10-year program. $10M–$20M/year. - **Elbit Border Surveillance:** 85–90% win probability. $20M CY2026 bookings. 14-year cycle. - **Golden Dome (Space Satellites):** Three aerospace primes in active design review. ~$16M per satellite payload. --- **The financials** Q3 FY2026: - Revenue: $19.1M — up **109% YoY** - Assemblies & Modules: up **355% YoY** - Gross margin: 36%, targeting 40% - Third consecutive positive adjusted EBITDA quarter - Record backlog: **$110.6M** — up 196% in one fiscal year - Cash: $55.2M — up from $4.9M nine months prior - GAAP net loss of $4.1M — almost entirely non-cash earnout revaluation Every quant screen sees a loss-making small-cap. The loss is a direct function of G5 booking $100M+ in orders — the better the business performs, the worse the income statement looks. That disconnect is the entry point. --- Priced as a commodity optics company. It is the domestic infrared monopoly the entire U.S. defense industrial base is legally mandated to use by 2030. The deadline doesn't move. The alternatives don't exist. --- *Not financial advice. Do your own due diligence.

English

14

31

267

28.8K

Vaelis@Vaelis_X·5m

Posted this two days ago. $HYLN up 28.50% today and fresh 52-week highs. The thesis is playing out faster than I expected. What the market is now starting to price in: revenue beat consensus by 146% last quarter, the Navy just named the USX-1 Defiant as the platform for KARNO sea trials — meaning military validation is now a named vessel with a timeline, not an abstract concept — and the live multi-fuel transition demo with zero power fluctuation removed the biggest technical risk bears were pointing to. The KARNO story was always about what happens when the Navy confirms it works under extreme conditions and commercial operators no longer have to take a technology risk. We might be getting closer to that moment than people realise. $139M cash. Zero debt. $400M LOI pipeline. Still under $1B market cap. Not financial advice — do your own diligence.

Vaelis@Vaelis_X

Bloom Energy just hit $74 billion market cap on a 2.8GW Oracle contract. The market is desperately searching for the next Bloom. It's already public. Under $850M market cap. Almost nobody has connected the dots. $HYLN (Hyliion Holdings) — the company the market still thinks is a failed EV truck startup. --- **The KARNO generator** In late 2022, Hyliion acquired KARNO directly from GE Aerospace's skunkworks for $37M — GE's own engineering team came with it, and GE became an equity stakeholder. KARNO is a Stirling-cycle linear heat generator. No combustion. No crankshaft. No spark plugs. No oil. A sealed working gas expands under external heat, driving a magnetic piston through stator coils to generate electricity. One moving part per shaft. Frictionless air bearings. Near-zero mechanical degradation. Decades of operational life. Maintenance: air filter replacements and remote software updates. Bloom's fuel cell stacks require full replacement every 3–5 years. KARNO doesn't. --- **Where KARNO beats Bloom outright** **Fuel agnosticism.** Bloom's SOFCs are poisoned by sulfur and carbon monoxide — requiring expensive external scrubbing on every installation. KARNO burns fuel externally to the sealed generator. Natural gas, hydrogen, biogas, flare gas, propane, ammonia, diesel — all validated. In-operation fuel switching demonstrated live in Q1 2026 without power fluctuation. **Dynamic load following.** Bloom must maintain ~800°C internally — ramping takes hours, forcing expensive battery buffers. KARNO modulates idle to full capacity in milliseconds via software-controlled stroke frequency. **Native 800VDC output.** Bloom outputs AC requiring conversion losses. KARNO natively outputs 800V DC — feeding directly into the emerging AI data center architecture standard. **Noise.** ~67 dB. Quieter than a conversation. Deployable inside data center corridors. --- **The manufacturing moat** KARNO's heat exchangers contain micro-channel geometries that cannot be cast or machined — only 3D printed in metal. The geometry itself unlocks commercial Stirling efficiency. Hyliion operates GE Colibrium M Line additive systems — the first commercial deployment in the United States — at 2–4x the throughput of standard industrial printers. Any competitor must bypass 105 patents across 22 families — including a flagship patent drawing 122 citations from Deere and Wabco — then master print parameters Hyliion and GE spent years optimizing. --- **Military validation** ~$20M in active Office of Naval Research contracts. Q1 2026: commenced building an 800 kW KARNO system for an unmanned Navy vessel. Targeting $40–50M in additional military contracts in 2026. If the Navy confirms KARNO runs reliably under extreme maritime conditions — commercial technology risk is eliminated. A data center operator buys with Navy-grade confidence. --- **The pipeline and financials** ~750 KARNO Cores under LOIs — over $400M in potential revenue. VFG Holdings (former Dell CTO/Chief AI Officer) signed for 250 Cores, 50MW over five years. Flexnode signed for edge AI. ABM Industries handles national field deployment. UL certification cleared early 2026. First 10 units delivering H2 2026. - Q1 2026 revenue: $2.83M — beat $1.15M consensus by 146% - 460% YoY growth - Cash: $139.3M. Zero debt. ~3-year non-dilutive runway - FY2026 guidance: ~$10M revenue, ~$100M year-end cash --- **The disconnect** Bloom Energy: $74B market cap. 26.7x revenue. Adjusted P/E of 10,861x. Hyliion: $832M market cap. $139M cash. ~$692M enterprise value. GE Aerospace-engineered platform. Native 800VDC. Millisecond load following. True fuel agnosticism. Near-zero degradation. Military-validated. $400M LOI pipeline. Priced as a failed SPAC. If 25% of LOIs convert over 12–18 months, the re-rating is violent. $3–5B is a modest mid-cap for this technology — 300–500% from here, with $139M as the floor. --- *Not financial advice. Do your own due diligence.

English

0

85

Vaelis@Vaelis_X·6m

Posted this yesterday. $LPTH up over 14.50% today, fresh 52-week high. This is exactly what happens when a thesis clicks. The GAAP loss was masking a company with revenue up 109% year over year, a $110M backlog up 196% in one fiscal year, and a legal mandate that every U.S. defense prime must use their glass by 2030. Canaccord just raised their target to $16.50. Analysts are now modeling $106M in revenue by 2027, up from $83M prior estimate. The market is starting to connect the dots. The 2030 NDAA deadline doesn't move. The alternatives still don't exist. BlackDiamond is the only NDAA-compliant, domestically produced infrared glass qualified at scale. Still watching this closely. Not financial advice — do your own diligence.

Vaelis@Vaelis_X

Congress just handed one company a legal monopoly on U.S. military optics. Almost nobody has noticed. $LPTH (LightPath Technologies) — ~$871M market cap. --- **The trigger** China controls the vast majority of global germanium refining — the material inside every U.S. military thermal imaging system. In 2023, China imposed export controls, driving a 65% price surge. Congress responded with the FY2026 NDAA: a hard January 1, 2030 deadline requiring complete elimination of adversary-sourced optical components from all U.S. military platforms. Every major defense prime must redesign their optical systems. One domestically produced, NDAA-compliant, germanium-free infrared glass is qualified at scale. LightPath's BlackDiamond™ — compounded entirely at ITAR-registered U.S. facilities. --- **Why BlackDiamond™ wins** Germanium's refractive index shifts with temperature, causing defocusing requiring heavy mechanical correction or cryogenic cooling. LightPath's BDNL-4™ — exclusively NRL-licensed — has a *negative* thermo-optic coefficient. Pair it with a positive dn/dT lens and thermal shifts cancel passively. No motors. No cooling. No weight penalty. BD6™ transmits across SWIR, MWIR, and LWIR in a single element — previously three separate systems. And unlike germanium, BlackDiamond™ is Precision Glass Moldable — high-volume, scalable, dramatically cheaper. --- **The moat** Exclusive rights to 14 chalcogenide glass compositions from the Naval Research Laboratory. No competitor can access them. - Phase 1: Three formulations qualified — already replacing germanium in active programs of record - Phase 2: Six more being qualified to MRL-9 — full high-rate production readiness Once qualified into a defense program, a material stays for the program's entire lifecycle. --- **Three acquisitions built the complete stack** **Visimid (2021):** Phoenix FPGA engine — turned LightPath from component supplier into systems integrator. **G5 Infrared (Feb 2025, $27.1M):** Long-range cooled MWIR cameras. G5 booked $100M+ in new orders since acquisition — triggering GAAP earnout revaluation, creating $12.2M in non-cash charges YTD that mask real profitability. **Amorphous Materials (Jan 2026, $7M):** Expanded maximum optic diameter from 5 to **17 inches** — unlocking space-based payloads. Already supplies glass for the F-35 targeting system and Apache Arrowhead sensor suite. --- **Defense programs already won** - **Lockheed NGSRI (Stinger Replacement):** Sole-sourced. Flight tests successful. 10,000 units/year at full rate. **$50M–$100M annual revenue** over 10 years. - **Air Force SEWADS (Counter-UAS):** $30M backlog. EdgeIR™ runs Hailo-8 AI at 26 TOPS — processes video locally, transmits only target coordinates. Unjammable. - **L3Harris Shipboard Threat Detection:** Sole-sourced. 10-year program. $10M–$20M/year. - **Elbit Border Surveillance:** 85–90% win probability. $20M CY2026 bookings. 14-year cycle. - **Golden Dome (Space Satellites):** Three aerospace primes in active design review. ~$16M per satellite payload. --- **The financials** Q3 FY2026: - Revenue: $19.1M — up **109% YoY** - Assemblies & Modules: up **355% YoY** - Gross margin: 36%, targeting 40% - Third consecutive positive adjusted EBITDA quarter - Record backlog: **$110.6M** — up 196% in one fiscal year - Cash: $55.2M — up from $4.9M nine months prior - GAAP net loss of $4.1M — almost entirely non-cash earnout revaluation Every quant screen sees a loss-making small-cap. The loss is a direct function of G5 booking $100M+ in orders — the better the business performs, the worse the income statement looks. That disconnect is the entry point. --- Priced as a commodity optics company. It is the domestic infrared monopoly the entire U.S. defense industrial base is legally mandated to use by 2030. The deadline doesn't move. The alternatives don't exist. --- *Not financial advice. Do your own due diligence.

English

0

58

Vaelis@Vaelis_X·1h

One overlooked UK-listed company is the critical hardware supplier behind SpaceX’s Starlink gateway network — and almost nobody is talking about it: $FTC (Filtronic plc) — £857M market cap It manufactures the E-band GaN Solid State Power Amplifiers that physically enable SpaceX’s Starlink gateway network. Not software. Not services. The actual hardware that closes the link budget between the ground and orbit. The Physical Problem LEO satellites orbit at 500–1,200km. Low altitude cuts latency to under 30ms but requires thousands of satellites. The real bottleneck isn’t the satellites — it’s the ground gateways backhauling orbital traffic to terrestrial fiber. At E-band frequencies (71–86 GHz), free-space path loss scales quadratically. You cannot close a multi-gigabit link budget over hundreds of kilometers of atmosphere without extreme transmit power. Silicon and GaAs physically cannot deliver it. GaN-on-SiC is the only answer — and Filtronic is the sole-source supplier for Starlink’s entire gateway network. Two Catalysts Just Collided SpaceX is targeting a $2T public valuation. Brookfield already accumulated a $2B pre-IPO stake. Roadshow starts June 4, 2026. The prospectus revealed Starlink did $11.4B in revenue in 2025 while the legacy launch business — $4.1B revenue — still posts net losses. Starlink IS the $2T valuation. And Starlink’s profitability scales directly with gateway expansion. More subscribers = more gateways = more Filtronic SSPAs. On April 30, 2026, the FCC voted 3-0 to scrap the 1990s-era EPFD power-limit regime — unlocking up to 7x more capacity for LEO operators. But you cannot realize 7x capacity with a software update. Every Starlink gateway on Earth needs physical hardware upgrades to high-power GaN SSPAs. The FCC ruling triggered a mandatory replacement cycle. Filtronic’s stock jumped double digits in a single session within a week of the vote. The Moat Filtronic’s proprietary “Cerus” SSPAs deliver 10–20W saturated output power in E-band with >20% Power Added Efficiency. Flip-chip packaging accurate to ±2µm on ceramic substrates. Competitors using conventional wire bonding exhibit high parasitic inductance at 80 GHz. That gap cannot be closed quickly. In April 2024, SpaceX signed a 5-year exclusive partnership and was granted warrants for up to 10% of Filtronic’s share capital — vesting in tranches tied directly to purchase order volume up to $19.7M cumulative spend. SpaceX earns equity as it buys more hardware. Competitors are structurally locked out. Tranches 1 and 2 already fully vested. Tranches 3 and 4 expected Q3/Q4 2026. The Numbers Space & Satcom revenue went from £2.1M in FY23 to £28.5M in FY25 to £58.4M projected FY26. Total revenue hits £78.2M projected. Gross margins expanding from 31.9% to 52.9%. Zero long-term debt. £18M projected net cash. Self-funded entirely from organic free cash flow. The Valuation Gap $GLW (Corning): 86.8x forward P/E, <5% space exposure, priced off one Nvidia fiber deal. $MTSI (MACOM): 38.2x forward P/E, <20% LEO exposure. $POET: unprofitable, retail meme volatility. $FTC: 16.5x forward P/E, 2.3x P/S, ~75% direct LEO revenue exposure, profitable, debt-free, sole-source SpaceX gateway supplier. Re-rate to a conservative 6x P/S on £78.2M FY26 revenue → £469M implied market cap → 150%+ upside before the SpaceX roadshow even opens. Before a single sovereign contract. Before Tranches 3 and 4 vest. Before US institutional money discovers it post-IPO. The physical backhaul of the entire space internet runs through one packaging line in Sedgefield, UK. SpaceX has 10% equity tied to keeping it that way. Not financial advice. DYOR.

English

0

1

177

Vaelis@Vaelis_X·18h

x.com/i/article/2057…

ZXX

0

355

Vaelis@Vaelis_X·1d

Broadcom just hit an $800 billion market cap on the AI data center infrastructure boom. The market is desperately searching for the next high-margin hardware monopoly. It's already public. Under $450M market cap. Almost nobody has connected the dots. $NVEC (NVE Corporation) — the company the market still thinks is a sleepy defense components contractor. The Spintronics platform NVE doesn't make traditional semiconductors. They are the pioneers of spintronics — technology that relies on the spin of electrons rather than just their electrical charge to store, process, and transmit data. Specifically, they manufacture Tunneling Magnetoresistance (TMR) and Giant Magnetoresistance (GMR) sensors and isolators. Instead of moving physical parts or relying on degrading light signals, NVEC's chips use microscopic magnetic fields. No optical degradation. Near-zero power consumption. Extreme sensitivity. Decades of operational life. Legacy silicon isolators and optocouplers wear out over time. Spintronics don't. Where Spintronics beats legacy silicon outright Traditional optocouplers rely on LEDs to transmit signals across isolated circuits, burning massive amounts of power and generating heat. NVE's spintronic isolators consume a fraction of the energy. In the power-hungry architectures of modern edge AI and EV battery management systems, thermal budgets are everything. Legacy optocouplers are also slow. NVE's spintronic couplers transmit data at gigabit speeds with nanosecond propagation delays — making them mission-critical for high-speed industrial robotics and automated data centers. Furthermore, LEDs inside traditional isolators dim and degrade over a few years, leading to system failures. Magnetic fields don't wear out. NVE's sensors boast an essentially infinite operational life, surviving in extreme temperatures and radiation where normal silicon melts or shorts. Spintronic micro-channel geometries allow NVEC to pack unprecedented sensitivity into wafer-level chip-scale packaging. The geometry itself unlocks commercial viability for next-gen autonomous systems. The manufacturing moat NVE is arguably the world's most profitable hardware company per capita. They have just 42 employees. Those 42 people generate over $15M in pure net profit annually. The barrier to entry is astronomical. Competitors would have to bypass decades of entrenched nanotechnology patents and specialized fabrication techniques. NVE is currently installing a new cluster of wafer fabrication equipment and pulling in advanced manufacturing tax credits, pivoting from pure R&D to a massive "Harvesting" phase. Their Q4 FY2026 gross margins just clocked in at 78%. Their net margin is 64%. You don't see 64% net margins unless you have an unbreakable monopoly on a highly specialized technology. Military and Enterprise validation NVE's tech is already military-validated. Their sensors are embedded in critical defense systems for anti-tamper and extreme aerospace environments. On the commercial side, their non-defense product sales just skyrocketed 34% YoY in Q4 2026. Their tech is deeply embedded in Abbott Laboratories' medical devices — proving the tech scales into mass-market hardware where miniaturization and zero-fail reliability are mandatory. If the U.S. military relies on it for defense systems, and the world's largest medical companies rely on it for human health — the commercial technology risk is eliminated. An industrial robotics or AI edge operator buys with ultimate confidence. The pipeline and financials Fiscal year 2026 proved the pivot to non-defense and industrial tech is working. Q4 2026 Net Income hit $4.93M, representing a massive 27% YoY increase. Gross margins reached 78% while net margins sat at 64%. The company holds roughly $43M in cash and marketable securities with zero debt. They even pay a steady 4.3% dividend yield ($1.00 per share quarterly payout) while you wait for the re-rating. For the full FY2026, they pulled in $26.3M in total revenue and converted $15.2M of that straight into net profit. The company is lean, aggressively profitable, and generating massive free cash flow per share. The disconnect Broadcom: $800B+ market cap. Trading at massive forward multiples. NVE Corporation: $434M market cap. ~$43M cash. ~$391M enterprise value. Proprietary spintronic nanotechnology. 64% net margins. Zero debt. Military and medical-grade validation. 4.3% dividend yield while you wait. Dominating a niche that is becoming critical for robotics, edge AI, and data center automation. Priced as a stagnant micro-cap value stock. If the market realizes spintronics is the bottleneck solution for next-gen robotics and low-power IoT over the next 12–18 months, the re-rating is violent. $1B to $1.5B is a modest mid-cap valuation for hardware monopolies with these margins — 200–300% from here, with a $43M cash fortress and a 4% yield as the floor. *Not financial advice. Do your own due diligence.

English

2

20

90

13K

Vaelis@Vaelis_X·1d

The energy layer is where most people anchor, but the actual gating constraint right now is high-voltage transformers and switchgear — lead times are still 2-3 years out and that's what's physically blocking interconnections, not the generation itself. Companies solving the physical grid tie-in problem are arguably the tightest bottleneck in the whole stack.

English

0

20

Rand Group@randgroup·2d

Behind every AI model there's a data center. Behind every data center there's a full stack of companies most investors never think about. This is the complete map with every ticker 👇 🔲 Compute Silicon This is where intelligence gets created. Every AI model, every training run, every inference call starts with a chip doing the math. The GPU narrative is well known by now, but the real evolution is happening in custom silicon. Hyperscalers are designing their own accelerators because no single vendor can serve every workload optimally. The companies building both general purpose GPUs and custom ASICs are sitting at the center of a market that's still in its early innings. Whoever owns the compute layer owns the foundation. Tickers: $NVDA $AMD $AVGO $INTC 🔲 Server OEMs & Solutions Chips are useless without systems to run them. This layer takes raw silicon and turns it into functioning infrastructure. AI server racks today are radically different from anything the industry built five years ago. Power density per rack has gone from 10kW to over 100kW. That means every aspect of the physical system, from liquid cooling to power distribution to rack architecture, has to be completely rethought. The companies solving these thermal and engineering challenges are extracting enormous value because without them, the chips can't operate at scale. Tickers: $SMCI $DELL $HPE $VRT $ETN $MOD 🔲 Memory & Storage AI models are only as good as the data they can access and how fast they can access it. High bandwidth memory has become the most supply constrained component in the entire stack. Training a frontier model requires moving enormous volumes of data in and out of the processor every millisecond, and traditional memory architectures can't handle it. The companies that locked in HBM production capacity early now have pricing power they haven't seen in decades. Storage matters too. The datasets feeding these models are growing exponentially, and the infrastructure to store, retrieve, and serve that data is a massive market on its own. Tickers: $SNDK $MU $WDC $P $NTAP 🔲 Networking & Connectivity The most underappreciated layer in the entire stack. You can fill a data center with the best GPUs on the planet, but if those chips can't talk to each other fast enough, performance collapses. AI training clusters require backend networking that's orders of magnitude faster than traditional cloud setups. Optical interconnects, high speed switches, and custom networking silicon are becoming the critical path for cluster performance. The spend flowing into this layer is accelerating because hyperscalers have realized that networking is the multiplier that makes their GPU investment actually productive. Tickers: $ANET $CSCO $MRVL $CRDO $CIEN $NOK 🔲 Neoclouds & Physical Infrastructure The hyperscalers can't build fast enough. That's created an opening for a new generation of infrastructure companies building GPU dense data centers and cloud platforms from the ground up. These are the picks and shovels of the next wave. They're signing long term contracts with enterprises that need compute now and can't wait two years for a hyperscaler slot. It's the most speculative layer, but the velocity of capital flowing into physical AI infrastructure is unlike anything the tech industry has seen before. Tickers: $NBIS $IREN $CRWV $APLD 🔲 Energy Everything above depends on this. A single frontier AI data center can draw as much power as a mid sized city, and the buildout is just getting started. Grid capacity is already strained in key markets. Nuclear is re entering the conversation as the only scalable, always on, zero carbon source that can meet the load. Natural gas is filling the gap while renewables scale. The companies providing generation, transmission, and grid solutions are the ultimate bottleneck play. If the energy doesn't scale, nothing else in this stack scales either. Tickers: $CEG $NEE $EOSE $GEV $EQT $VST The AI trade is a full stack trade. The winners of this cycle won't just be the ones making the processors. They'll be the ones cooling them, connecting them, powering them, and housing them. Understanding these six layers is how you build real exposure to the AI infrastructure buildout instead of just chasing the obvious names. Which layer are you most focused on?

English

29

162

718

56.5K

Vaelis@Vaelis_X·1d

@adamtlr @Longviewres Yes I have, I posted about it not too long ago and for me they are so unreasonably mispriced, great article from @Longviewres though, that’s where my first conviction in it stemmed from

English

0

1

1.7K

No Amygdala@adamtlr·1d

@Vaelis_X Have you looked at $CGEH? If not you should read @Longviewres super detailed article about it.

English

1

0

3

1.9K

Vaelis@Vaelis_X·2d

Bloom Energy just hit $74 billion market cap on a 2.8GW Oracle contract. The market is desperately searching for the next Bloom. It's already public. Under $850M market cap. Almost nobody has connected the dots. $HYLN (Hyliion Holdings) — the company the market still thinks is a failed EV truck startup. --- **The KARNO generator** In late 2022, Hyliion acquired KARNO directly from GE Aerospace's skunkworks for $37M — GE's own engineering team came with it, and GE became an equity stakeholder. KARNO is a Stirling-cycle linear heat generator. No combustion. No crankshaft. No spark plugs. No oil. A sealed working gas expands under external heat, driving a magnetic piston through stator coils to generate electricity. One moving part per shaft. Frictionless air bearings. Near-zero mechanical degradation. Decades of operational life. Maintenance: air filter replacements and remote software updates. Bloom's fuel cell stacks require full replacement every 3–5 years. KARNO doesn't. --- **Where KARNO beats Bloom outright** **Fuel agnosticism.** Bloom's SOFCs are poisoned by sulfur and carbon monoxide — requiring expensive external scrubbing on every installation. KARNO burns fuel externally to the sealed generator. Natural gas, hydrogen, biogas, flare gas, propane, ammonia, diesel — all validated. In-operation fuel switching demonstrated live in Q1 2026 without power fluctuation. **Dynamic load following.** Bloom must maintain ~800°C internally — ramping takes hours, forcing expensive battery buffers. KARNO modulates idle to full capacity in milliseconds via software-controlled stroke frequency. **Native 800VDC output.** Bloom outputs AC requiring conversion losses. KARNO natively outputs 800V DC — feeding directly into the emerging AI data center architecture standard. **Noise.** ~67 dB. Quieter than a conversation. Deployable inside data center corridors. --- **The manufacturing moat** KARNO's heat exchangers contain micro-channel geometries that cannot be cast or machined — only 3D printed in metal. The geometry itself unlocks commercial Stirling efficiency. Hyliion operates GE Colibrium M Line additive systems — the first commercial deployment in the United States — at 2–4x the throughput of standard industrial printers. Any competitor must bypass 105 patents across 22 families — including a flagship patent drawing 122 citations from Deere and Wabco — then master print parameters Hyliion and GE spent years optimizing. --- **Military validation** ~$20M in active Office of Naval Research contracts. Q1 2026: commenced building an 800 kW KARNO system for an unmanned Navy vessel. Targeting $40–50M in additional military contracts in 2026. If the Navy confirms KARNO runs reliably under extreme maritime conditions — commercial technology risk is eliminated. A data center operator buys with Navy-grade confidence. --- **The pipeline and financials** ~750 KARNO Cores under LOIs — over $400M in potential revenue. VFG Holdings (former Dell CTO/Chief AI Officer) signed for 250 Cores, 50MW over five years. Flexnode signed for edge AI. ABM Industries handles national field deployment. UL certification cleared early 2026. First 10 units delivering H2 2026. - Q1 2026 revenue: $2.83M — beat $1.15M consensus by 146% - 460% YoY growth - Cash: $139.3M. Zero debt. ~3-year non-dilutive runway - FY2026 guidance: ~$10M revenue, ~$100M year-end cash --- **The disconnect** Bloom Energy: $74B market cap. 26.7x revenue. Adjusted P/E of 10,861x. Hyliion: $832M market cap. $139M cash. ~$692M enterprise value. GE Aerospace-engineered platform. Native 800VDC. Millisecond load following. True fuel agnosticism. Near-zero degradation. Military-validated. $400M LOI pipeline. Priced as a failed SPAC. If 25% of LOIs convert over 12–18 months, the re-rating is violent. $3–5B is a modest mid-cap for this technology — 300–500% from here, with $139M as the floor. --- *Not financial advice. Do your own due diligence.

English

21

50

373

63.8K

Vaelis@Vaelis_X·1d

The $35B → $100B HBM TAM projection is the right framing — Citigroup just raised their $MU price target to $840 today on exactly this thesis. Micron's HBM3E is qualifying into Nvidia's Blackwell supply chain alongside SK Hynix, making this a 2-horse race at the most critical bottleneck in AI compute.

English

1

0

2

238

Qualtrim@qualtrim·3d

Micron is seeing massive structural demand for their high bandwidth memory. The HBM market was worth $35B in 2025. Micron projects this TAM to grow to $100B by 2028, suggesting a wild 40% CAGR. Reminder: 7x forward multiple. $MU

English

19

21

295

16.9K

Vaelis@Vaelis_X·1d

Three months from design to batch production is the stat that matters most here — traditional industrial robotics takes 2-3 years for that cycle. The supply chain constraint will quickly shift to specialized actuators and precision force sensors, which is where the next semiconductor-adjacent bottleneck is forming.

English

0

1

24

Lukas Ziegler@lukas_m_ziegler·3d

Generative Bionics designed a humanoid robot from scratch and sent it to batch production. They did it in THREE MONTHS. 🇮🇹 The GENE01 from @G_Bionics ships with: → Two scalable lower body configurations → Physical AI deployed for motor control and world-action modelling → Blind walking and running enabled entirely by reinforcement learning → RL that generalises across different robot platforms out of the box Nice to see walking robots so fast after announcing the round :)

Lukas Ziegler@lukas_m_ziegler

🚨 BREAKING: Italy just entered the humanoid race in a serious way! 🇮🇹 Generative Bionics, a new spin-out from the Italian Institute of Technology has raised €70M to build “Made in Italy” humanoids designed for real industrial, healthcare, and retail work. The operation is led by CDP Venture Capital SGR's Artificial Intelligence Fund, with the participation of AMD Ventures, Duferco, Eni Next, RoboIT and Tether. It’s one of the largest deep-tech rounds in Europe for humanoids. The company is led by Daniele Pucci and brings over 20 years of IIT robotics research into one stack: – distributed tactile + force sensing from iCub – Physical AI architectures from ergoCub – robust control + simulation methods from iRonCub (the flying humanoid) The goal is simple, but very challanging: humanoids that can operate safely around people and handle repetitive, high-risk, or labor-intensive tasks, with real contracts already in the pipeline for 2026. The new funding will accelerate product development, Physical AI training, industrial validation, and the build-out of their first manufacturing plant. According to their projections, the humanoid market could pass $200B by 2035 and reach the trillions by mid-century. The first full robot will debut at CES 2026 in Las Vegas. Good moment for European robotics! 🇪🇺 ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com

English

11

43

256

26.9K

Vaelis@Vaelis_X·1d

What jumps out is SK Hynix at $951B — they're the dominant HBM supplier and yet most semiconductor discussions focus on GPU makers. No HBM, no H100/H200/Blackwell at scale. SK Hynix controls ~50% of that market with Micron just beginning to ramp, making memory the real choke point in AI infrastructure.

English

0

135

Global Statistics@Globalstats11·2d

Biggest Semiconductor Companies in the World 💻 1. 🇺🇸 NVIDIA - $5.4 Trillion 2. 🇹🇼 TSMC - $2.1 Trillion 3. 🇺🇸 Broadcom - $2.0 Trillion 4. 🇰🇷 Samsung - $1.3 Trillion 5. 🇰🇷 SK hynix - $951 Billion 6. 🇺🇸 Micron - $864 Billion 7. 🇺🇸 AMD - $731 Billion 8. 🇺🇸 Intel - $606 Billion 9. 🇳🇱 ASML - $586 Billion 10. 🇺🇸 Lam Research - $362 Billion 11. 🇺🇸 Applied Materials - $342 Billion 12. 🇺🇸 Texas Instruments - $269 Billion Source: CompaniesMarketCap

English

8

136

507

34.6K

Vaelis@Vaelis_X·1d

The transmission line overlay is what most investors miss — the actual bottleneck isn't real estate or water, it's high-voltage interconnection capacity. Power companies already on existing 500kV/345kV infrastructure have a permitting moat that new entrants simply can't replicate in under a decade.

English

0

22

Jack Prandelli@jackprandelli·2d

The US Department of Energy just mapped every data center in America. This is what the AI power grid looks like. The dots are data centers. Yellow = operating. Orange = under construction. White = planned. The lines are high-voltage transmission 735kV, 500kV, 345kV the arteries that move electrons from generators to compute loads. Look at the density along the East Coast, Northern Virginia to the Carolinas. Then look at Texas. Then Northern California. The largest circles on this map represent facilities demanding over 5,000 MW of power. Single campuses pulling more electricity than mid-sized cities. Northern Virginia is so dense the dots overlap. Data centers cluster on transmission corridors. Not because land is cheap because power is available. When the line is full, the next data center goes somewhere else. The grid is the bottleneck. Every orange dot is a power purchase agreement being negotiated right now. Every white dot is a utility commission filing, a gas plant approval, a pipeline capacity booking. The $66.8 bn NextEra-Dominion deal, Meta's 10 new gas plants in Louisiana, the Alaska LNG FID push they all trace back to maps that look like this. AI infrastructure is built in substations, on transmission corridors, and at the end of gas pipelines. Link in the comments, to see my stocks 👇

English

341

2.7K

4.4K

324.4K

Vaelis@Vaelis_X·1d

@MarioNawfal Force control and weight estimation is what separates Atlas here. The real investor angle: physical AI models like this run on Nvidia's Isaac GR training stack, which means humanoid robotics is a new durable GPU demand source that almost no analyst model is pricing in yet.

English

0

15

Mario Nawfal@MarioNawfal·3d

🇺🇸 Atlas can now lift heavy objects using full-body AI coordination. Boston Dynamics' humanoid robot just cleared one of the hardest physical barriers in robotics: making a machine move like it actually understands weight.

English

61

120

533

56.3K

Vaelis@Vaelis_X·1d

The historical analog worth flagging: the telecom buildout ran the hardest in the final 18 months of the 1999-2000 cycle. If we're in late stage, AI infrastructure capex could still rip before the turn — the question is whether you can exit the picks-and-shovels plays before the hyperscaler spending reversal.

English

0

1

20

Rand Group@randgroup·3d

What if 2026 is not the beginning of the bull market, but the final stage of it?

English

36

25

219

38.1K

Vaelis@Vaelis_X·1d

@PolymarketMoney What's underappreciated here is that Vera is a CPU, not a GPU — it means Nvidia is no longer just the accelerator company. Hyperscalers are now locked into the Nvidia stack at the host processor level too, which deepens the moat significantly ahead of tonight's earnings.

English

0

48

Polymarket Money@PolymarketMoney·3d

$NVDA delivered its first Vera CPUs to Anthropic, OpenAI, SpaceX and Oracle Cloud.

English

11

19

254

11.2K

Vaelis@Vaelis_X·1d

What's different this time is Nvidia just shipped its first Vera CPUs to Anthropic, OpenAI, SpaceX and Oracle Cloud right before the print — that's a structural product catalyst that didn't exist in prior earnings cycles. The macro headwinds matter less when you're actively delivering next-gen AI infrastructure at scale.

English

0

1

153

Ted@TedPillows·2d

This is very interesting. Nvidia $NVDA will report its earnings today. Last time they reported their earnings, it marked the local top. But before that, Nvidia earnings marked the local bottom. This time, markets are already looking weak with yields soaring and oil above $100. Will Nvidia earnings mark the local bottom or the stock market correction will continue?

English

120

69

421

55.7K

Vaelis@Vaelis_X·1d

The most important earnings call in AI history drops tonight. $NVDA after close. $78.8B revenue expected. 21 beats in the last 23 quarters. But the number everyone is actually watching is Q2 guidance. Consensus $85-87B. Whisper $90B. Jensen speaks at 5PM ET. The transcript matters more than the chart tonight.

English

0

556

Vaelis@Vaelis_X·2d

I’ll be watching the call at 5PM ET. The transcript matters more than the chart tonight.

English

0

229

Vaelis@Vaelis_X·2d

BofA has a $320 price target. Wedbush is pointing to a trillion dollar order book through 2027. The setup isn’t whether Nvidia beats — it’s whether Jensen raises the ceiling on what 2027 looks like.

English

1

0

270

Vaelis@Vaelis_X·2d

$NVDA reports in a few hours. Everyone is watching the revenue number. I’m watching something else entirely. Here’s what actually matters tonight. 👇

English

1

0

294

Vaelis

Keşfet