🚀 Lumentum just dropped one of the strongest signals of the year on AI optical infrastructure. CEO Michael Hurlston (Bloomberg interview today): optical components are “completely sold out” through early 2029. Demand from hyperscalers is so intense that Lumentum can’t meet ~70% of demand for some products through next year, and still faces a ~30% shortfall on others into 2028. Customers are already locking orders for 2030. This is the “nervous system” of next-gen AI clusters — high-speed optical interconnects (InP devices, EMLs, pump lasers). | Metric | Details | Impact | |---------------------------------|--------------------------------------|-------------------------------------| | Sold-out horizon | Through early 2029 | Visibility locked | | Unmet demand (some products) | ~70% through 2027 | Severe supply crunch | | Unmet demand (other products) | ~30% through 2028 | Bottleneck continues | | Nvidia relationship | $2B investment + purchase agreements | Strategic alignment | | Capacity expansion | Japan fabs + considering M&A | Trying to catch up | My take: Extremely bullish for the optical layer of AI infrastructure. When a Nvidia-backed photonics player says there’s “no end in sight,” it confirms interconnects are becoming as critical as the GPUs themselves. Do you think optical interconnects (silicon photonics / InP) will become the next major bottleneck — or the next big winner — in the AI supercycle? Drop your thoughts 👇 @Lumentum @NVIDIA #AIInfrastructure #Semiconductors #DataCenters Source (verified today – Bloomberg 9 Oct 2026): bloomberg.com/news/articles/… --- 🚀 Lumentum vient de balancer l’un des signaux les plus forts de l’année sur l’infrastructure optique de l’IA. Le CEO Michael Hurlston (interview Bloomberg aujourd’hui) : les composants optiques sont « complètement sold out » jusqu’au début 2029. La demande des hyperscalers est tellement intense que Lumentum ne peut pas satisfaire ~70 % de la demande sur certains produits jusqu’à l’année prochaine, et fait encore face à un manque d’environ 30 % jusqu’en 2028. Les clients commencent déjà à verrouiller des commandes pour 2030. C’est le « système nerveux » des clusters IA de nouvelle génération — les interconnexions optiques haute vitesse (composants InP, EML, pump lasers). Mon avis : extrêmement bullish pour la couche optique de l’infrastructure IA. Quand un acteur photonique soutenu par Nvidia dit qu’il n’y a « pas de fin en vue », ça confirme que les interconnexions deviennent aussi critiques que les GPUs eux-mêmes. Pensez-vous que les interconnexions optiques (silicon photonics / InP) vont devenir le prochain grand bottleneck (ou le prochain grand gagnant) du supercycle IA ? Donnez votre avis 👇

🚨 CoreWeave just dropped the strongest signal of the year on AI infrastructure. Q2 2026: • Revenue +112% YoY → $2.6B • Backlog → **$104.2 billion** • 2026 Capex raised to **$35-39 billion** Stock surged ~20%. Near-term capacity = sold out. AI cloud demand isn’t slowing down. It’s accelerating. --- 🚨 CoreWeave vient de balancer le signal le plus fort de l’année sur l’IA infrastructure. Q2 2026 : • Revenue +112 % YoY → $2.6B • Backlog → **$104.2 milliards** • Capex 2026 relevé à **$35-39 milliards** Le stock a explosé de ~20 %. Near-term capacity = sold out. La demande AI cloud ne ralentit pas. Elle s’accélère. --- Key figures / Chiffres clés : | Metric / Métrique | Q2 2026 | Impact | |-----------------------|------------------|---------------------------------| | Revenue | $2.6B (+112 %) | Record | | Backlog | $104.2B | ~8 years of current revenue | | Capex 2026 | $35-39B | +$4B vs previous guidance | | Active power | 1.5 GW | +500 MW in one quarter | | New commitments Q3 | >$25B | Massive visibility | CoreWeave is building AI factories at a pace even hyperscalers struggle to match. CoreWeave construit des usines d’IA à une vitesse que même les hyperscalers peinent à suivre. --- This is no longer just “more Nvidia GPUs”. It’s a **new financial model**: neoclouds are massively leveraging debt to turn compute into an investable asset class (Nvidia is pushing a $500B third-party capital plan with BlackRock, Blackstone, etc.). Ce n’est plus juste « plus de GPUs Nvidia ». C’est un **nouveau modèle financier** : les neoclouds s’endettent massivement pour transformer le compute en asset class investissable (Nvidia pousse un plan de $500B de capital tiers avec BlackRock, Blackstone, etc.). Meanwhile, the bottlenecks are shifting: Pendant ce temps, les goulots d’étranglement se déplacent : → Power (48V saturates → 800V DC) → Cooling (Wiwynn already testing diamond composite) → Packaging & networking AI infrastructure is entering its “capital-intensive + extreme engineering” phase. L’infrastructure IA entre dans sa phase « capital intensif + engineering extrême ». --- Are CoreWeave (and the neoclouds) becoming the real winners of 2026-2027… or will the massive debt + capex leverage create the next fragility point in the AI supercycle? Est-ce que CoreWeave (et les neoclouds) sont en train de devenir les vrais winners de 2026-2027… ou est-ce que le levier de dette + capex monstrueux va créer le prochain point de fragilité du supercycle AI ? Your take / Votre take 👇 #AIInfrastructure #CoreWeave #DataCenters #Semiconductors #Nvidia

📊📊 IREN Update (NASDAQ: IREN) – July 30, 2026 Closing price: $29.31 (-13.6%) After-hours ≈ $29.90 Market cap ≈ $10.5B 52-week range: $14.72 – $76.87 Performance: • 1 month: ≈ -40% • 1 year: +80% to +100% The company: Former Bitcoin miner (ex-Iris Energy) aggressively pivoting to AI Cloud with 100% renewable-powered data centers. Major news (July 20): → Signed $2.8 billion in multi-year AI contracts → 2026 AI Cloud ARR target raised to over $4B (85% already under contract) → Clients: Microsoft, NVIDIA, Perplexity, etc. → Planned capacity: 480 MW by end of 2026 → 1.2 GW in 2027 (multi-GW ambition) Why the drop? Sector-wide pressure on AI / data center stocks + elevated short interest (≈ 28%) + high valuation concerns. No company-specific bad news. Outlook: • Strong revenue growth expected • Analyst consensus: Buy • Average price target ≈ $82 (high $105-131, low ≈ $46) → Theoretical upside > 150% Solid growth story with good visibility, but highly volatile stock. What do you think — is this a buying opportunity or still too risky right now? 👀 #IREN #AI #Bitcoin

1/8 🚨 China’s AI Champion is Building Its Own Chips DeepSeek, the Chinese startup behind some of the world’s most popular AI models, is now developing its own AI chip, according to sources. The goal? Reduce its heavy reliance on Nvidia and Huawei chips. 2/8 Why does this matter? DeepSeek has been one of the biggest success stories in Chinese AI. Until now, it has largely depended on foreign (mainly Nvidia) and Huawei chips to train and run its models. Moving into chip design marks a major strategic shift. 3/8 Key points from the report: • DeepSeek is actively hiring chip design engineers • The move aims to reduce dependence on Nvidia GPUs and Huawei chips • If successful, it would make DeepSeek far more independent in the AI race This is not just another startup story — it’s China’s AI flagship making a big bet on vertical integration. 4/8 This fits into a bigger trend we’re seeing in 2026: Hyperscalers and top AI labs are increasingly moving toward **custom silicon** to gain more control over performance, cost, and supply chain. DeepSeek joining this movement shows that even non-chip companies feel the pressure to control their own hardware destiny. 5/8 What could this mean long term? • More competition in the AI chip space (beyond Nvidia, AMD, and Huawei) • Potential acceleration of China’s domestic semiconductor ecosystem • Increased focus on custom architectures (including chiplets) to optimize for specific workloads The AI hardware landscape is becoming more fragmented — and more interesting. 6/8 For the global semiconductor industry, this is another signal that AI demand is forcing structural changes. Companies that can offer flexible, high-performance solutions (whether through advanced packaging, chiplets, or custom design) are likely to benefit from this shift. 7/8 Short-term noise around AI spending doesn’t change the bigger picture: the race for AI leadership is also becoming a race for **hardware sovereignty**. DeepSeek’s move is one more piece of that puzzle. 8/8 What do you think? Will we see more AI companies follow DeepSeek and develop their own chips, or will most continue relying on Nvidia and custom solutions from big players? Drop your thoughts 👇 #Semiconductors #AI #DeepSeek #ChinaTech #CustomSilicon

1/8 🚨 AI Chip Stocks Just Got Hammered Micron plunged **-13%**, Intel **-9%**, AMD **-7%**. The semiconductor sector is suddenly questioning whether the massive AI infrastructure spending can continue at this pace beyond 2026. Is the party over… or is this just a healthy reset? 2/8 What triggered this sharp sell-off? Main reasons: • SK Hynix reportedly slowing High-Bandwidth Memory (HBM) expansion • Growing doubts about real ROI from hyperscalers’ huge AI capex • More hawkish signals from the Federal Reserve After months of euphoria, the market is now demanding proof that all this spending will actually pay off. 3/8 Quick breakdown of the situation: | Factor | Impact | Why it matters | |-------------------------------|---------------------------------|---------------------------------------------| | SK Hynix slowing HBM | Memory & AI chip stocks hit hardest | HBM is the key bottleneck for AI training & inference | | AI capex sustainability fears | Broad sell-off across semis | Hyperscalers may slow down if returns look weak | | Hawkish Fed | Higher rates pressure growth stocks | Makes big AI projects more expensive to finance | | Previous massive rally | Valuations looked stretched | Any negative news triggers violent corrections | 4/8 This isn’t the death of the AI story. It’s a **necessary correction** after an extremely fast run-up. The market is forcing the industry to focus on **real efficiency and returns** instead of unlimited spending. 5/8 **The unexpected twist**: This volatility actually **accelerates the importance of chiplet architectures**. Chiplets offer: • Higher manufacturing yields • Lower costs • Better performance per watt • Faster design cycles When every dollar of AI investment must be justified, **modular chiplet designs become a major competitive advantage**. 6/8 Long-term outlook stays bullish. The AI infrastructure buildout will continue, but it will become **more selective, more efficient, and increasingly chiplet-driven**. Companies that master advanced packaging and chiplet integration will be the big winners in the next phase. 7/8 Short-term pain for traders… but this kind of reset is exactly what separates hype from real technological progress. Chiplets aren’t just surviving this environment — they’re becoming **essential**. 8/8 What do you think? Is this July 2026 sell-off a healthy reality check for AI spending… or the start of a bigger slowdown? Drop your thoughts below 👇 #Semiconductors #AI #Chiplets #AIInfrastructure

🚀 Chiplet technology is becoming one of the most important advancements in the semiconductor industry, especially for AI applications. Instead of relying on a single massive chip, companies are now connecting multiple smaller specialized chips (chiplets) within the same package. This approach allows for better performance, improved manufacturing yields, and more flexible designs — particularly useful for building powerful AI accelerators. Major players are increasingly adopting chiplet architectures to overcome the physical and economic limits of traditional monolithic chip designs. | Metric | Details | Impact | |---------------------------|------------------------------------------------------|-----------------------------------------------------| | Design Approach | Multiple specialized chiplets instead of one large die | Better performance and design flexibility | | Manufacturing Yield | Smaller dies are easier and cheaper to produce | Reduces costs and improves production efficiency | | AI Performance | Enables higher compute density for AI workloads | Supports more powerful and scalable AI systems | | Scalability | Easier to mix different process nodes in one package | Faster innovation and customization | Chiplets represent a fundamental shift in how advanced chips are designed and built. Rather than trying to make one giant chip that becomes increasingly difficult and expensive to manufacture, the industry is moving toward modular architectures. This not only helps with current AI demands, but also opens the door to faster iteration and more specialized processors in the future. It’s one of the smartest engineering responses to the growing complexity of modern computing. This is a **positive development** for the semiconductor and AI industries. Chiplet technology is helping the sector continue scaling performance despite the increasing challenges of advanced manufacturing. Do you think chiplet architectures will become the standard for high-performance AI chips in the coming years? #Semiconductors #AI #Chiplets Source: Recent advancements in chiplet technology (July 2026) → anandtech.com/ --- 🚀 La technologie des chiplets devient l’une des avancées les plus importantes dans l’industrie des semi-conducteurs, notamment pour les applications d’IA. Au lieu de s’appuyer sur une seule puce massive, les entreprises connectent désormais plusieurs petites puces spécialisées (chiplets) au sein d’un même boîtier. Cette approche permet d’obtenir de meilleures performances, des rendements de fabrication améliorés et des conceptions plus flexibles — particulièrement utile pour construire des accélérateurs d’IA puissants. Les grands acteurs adoptent de plus en plus les architectures chiplet pour contourner les limites physiques et économiques des conceptions de puces monolithiques traditionnelles. | Métrique | Détails | Impact | |---------------------------|------------------------------------------------------|-----------------------------------------------------| | Approche de conception | Plusieurs chiplets spécialisés au lieu d’une seule grande puce | Meilleure performance et flexibilité de conception | | Rendement de fabrication | Les petites puces sont plus faciles et moins chères à produire | Réduit les coûts et améliore l’efficacité de production | | Performance IA | Permet une densité de calcul plus élevée pour les charges IA | Soutient des systèmes d’IA plus puissants et scalables | | Scalabilité | Plus facile de mixer différents nœuds de fabrication dans un même boîtier | Accélère l’innovation et la personnalisation | Les chiplets représentent un changement fondamental dans la façon dont les puces avancées sont conçues et fabriquées. Plutôt que d’essayer de fabriquer une seule puce géante qui devient de plus en plus difficile et coûteuse à produire, l’industrie se tourne vers des architectures modulaires. Cela permet non seulement de répondre aux besoins actuels de l’IA, mais aussi d’ouvrir la voie à une itération plus rapide et à des processeurs plus spécialisés dans le futur. C’est l’une des réponses d’ingénierie les plus intelligentes face à la complexité croissante du calcul moderne. C’est un **développement positif** pour les industries des semi-conducteurs et de l’IA. La technologie des chiplets aide le secteur à continuer d’améliorer les performances malgré les défis croissants de la fabrication avancée. Pensez-vous que les architectures chiplet deviendront la norme pour les puces IA haute performance dans les années à venir ?

Haha bonne analyse… Où va le Bitcoin finalement ? Personne ne le sait vraiment 😏🚀 Haha good analysis… Where is Bitcoin heading in the end? Nobody really knows 😏🚀

🚀 Bitcoin continues to build its institutional foundation even during periods of short-term volatility. Despite recent market fluctuations, Bitcoin’s core adoption story remains strong. Institutions and long-term holders continue to recognize Bitcoin as a superior store of value and a hedge against traditional financial system risks. The infrastructure around Bitcoin (ETFs, corporate treasuries, and payment rails) keeps expanding quietly but steadily. This resilience shows that Bitcoin is maturing beyond short-term price action and is increasingly viewed as a strategic asset by serious capital. | Metric | Details | Impact | |---------------------------|------------------------------------------------------|-----------------------------------------------------| | Long-term Conviction | Institutions and corporates continue holding through volatility | Strengthens Bitcoin’s position as a reserve asset | | Infrastructure Growth | Ongoing development of ETFs, custody solutions, and payment systems | Makes Bitcoin more accessible and usable at scale | | Market Maturity | Bitcoin increasingly seen as “digital gold” by traditional finance | Reduces reliance on retail-driven price action | | Adoption Signal | Continued interest from serious capital despite noise | Builds a stronger foundation for the next cycle | Short-term volatility often distracts from what really matters. While price action can be noisy, the quiet, consistent building of Bitcoin’s institutional infrastructure is what will matter most over the next 3–5 years. The fact that serious players keep accumulating and developing around Bitcoin, even when sentiment is mixed, is one of the strongest bullish signals for the asset’s long-term future. @saylor has been one of the most consistent voices highlighting this long-term vision. This is a **positive reminder** of Bitcoin’s growing maturity. The foundation is being laid brick by brick, and those with a long-term view are positioning accordingly. Do you think Bitcoin’s long-term adoption story is stronger than the current market noise suggests? #Bitcoin #Crypto #BitcoinAdoption Source: Market data & institutional reports (June 29, 2026) → coinstats.app/ai/a/latest-ne… --- 🚀 Bitcoin continue de construire ses fondations institutionnelles même pendant les périodes de volatilité à court terme. Malgré les fluctuations récentes du marché, le récit d’adoption de Bitcoin reste solide. Les institutions et les détenteurs long terme continuent de voir le Bitcoin comme une réserve de valeur supérieure et une protection contre les risques du système financier traditionnel. L’infrastructure autour du Bitcoin (ETFs, trésoreries corporate, solutions de paiement) continue de se développer de façon discrète mais constante. Cette résilience montre que Bitcoin mûrit au-delà des mouvements de prix à court terme et est de plus en plus considéré comme un actif stratégique par les capitaux sérieux. | Métrique | Détails | Impact | |---------------------------|------------------------------------------------------|-----------------------------------------------------| | Conviction long terme | Institutions et entreprises continuent de détenir malgré la volatilité | Renforce la position du Bitcoin comme actif de réserve | | Croissance de l’infrastructure | Développement continu des ETFs, solutions de custody et systèmes de paiement | Rend Bitcoin plus accessible et utilisable à grande échelle | | Maturité du marché | Bitcoin de plus en plus vu comme « l’or numérique » par la finance traditionnelle | Réduit la dépendance aux mouvements retail | | Signal d’adoption | Intérêt continu des capitaux sérieux malgré le bruit | Construit des fondations plus solides pour le prochain cycle | La volatilité à court terme fait souvent oublier l’essentiel. Pendant que le prix fait du bruit, la construction silencieuse et constante de l’infrastructure institutionnelle de Bitcoin est ce qui comptera le plus dans les 3 à 5 prochaines années. Le fait que des acteurs sérieux continuent d’accumuler et de développer autour du Bitcoin, même quand le sentiment est mitigé, est l’un des signaux haussiers les plus forts pour l’avenir à long terme de l’actif. @saylor fait partie des voix les plus constantes à défendre cette vision long terme. C’est un **rappel positif** de la maturité croissante de Bitcoin. Les fondations sont posées brique par brique, et ceux qui ont une vision long terme se positionnent en conséquence. Pensez-vous que le récit d’adoption long terme de Bitcoin est plus fort que le bruit actuel du marché ? #Bitcoin #Crypto #BitcoinAdoption Source : Données de marché & rapports institutionnels (29 juin 2026) → coinstats.app/ai/a/latest-ne…