Your signal. Your price.
Nvidia has committed to guaranteeing $250 billion in financing for OpenAI's 10-gigawatt data center project, which could cost over $500 billion. This guarantee, representing 135% of Nvidia's retained earnings, significantly increased Nvidia's credit risk and negatively impacted its stock, according to Jim Chanos.
Kimmy K3 shows better output token efficiency than many Anthropic models and Opus 5 in some high-reasoning benchmarks, despite overall token hunger. However, running the trillion-parameter model locally demands substantial hardware, requiring 64 H100 GPUs.
Marty Bent notes the AI dominance race includes debates on open-source versus closed-source models, citing Scott Bessent's stance on sanctioning China for AI model theft. John argues the US administration primarily seeks to limit China's ability to "free ride" on frontier AI through distillation attacks and circumvention of chip export controls, rather than opposing open-source models generally.
John highlights the formation of the Open Secure AI Alliance by NVIDIA and partners, which includes infrastructure and chip companies like Synopsys, Cadence, Cisco, Cloudflare, CrowdStrike, IBM, HP, and Dell. This alliance demonstrates that many American tech companies advocate for a more open and broadly available AI infrastructure, rather than a system controlled by an oligopsony of two frontier labs.
Dario Amodei identifies hardware and industrial-scale distillation as critical leverage points for AI security. He advocates for tighter export controls on high-end chips to maintain a strategic lead.
Dario Amodei believes a model secretly trained by the CCP for military use poses a greater national security threat than any public open-weights release. Choking chip supply is key.
Government "bank shots," such as banning chips, are indirect diplomatic tools in trade wars, designed to impact competitors without directly addressing the core issue.
Ben Horowitz explains that Chinese labs have outpaced American counterparts in open-source AI performance because U.S. labs built proprietary models, necessitating high valuations to acquire expensive GPUs, and historically, market leaders rarely adopt open-source strategies.
This new Nvidia initiative targets smaller, less established firms. Fermis plans to deploy a cluster of 170,000 GPUs in Indonesia, while Sharon AAI aims for 40,000 GB300 GPUs, making them the first NeoClouds under this program.
Rich Dupri of 247 Wall Street explains that Nvidia's strategy differs from dot-com era vendor financing because the current bottleneck is capital for AI factories, not demand. Nvidia provides anchor demand to unlock external financing without taking on direct default risk, operating with quarterly revenues of $80 billion.
Kazu Nakashigi's company, Yakumo, builds quantum computer hardware, specifically Quantum Processing Units (QPUs), which are difficult to create and require many units for useful calculations.
Kazu Nakashigi notes that quantum computing has been '5 to 10 years away' for 30 years primarily due to hardware challenges, making it hard to create special corrugated units (QPUs).
Kazu Nakashigi clarifies a misconception: quantum computers are not universally faster than CPUs or GPUs, but are applied to specific difficult problems like combinations or RSA cryptography.
The required number of qubits for practical quantum computing has significantly decreased from one million to approximately 10,000, while physical qubit counts have increased from 1-10 to 6,000.
Kazu Nakashigi predicts that the necessary and achievable qubit counts will converge around 2030, enabling more widespread real-world applications of quantum computers in data centers, with some already installed in Japan by companies like Toyota.
Quantum calculations currently make many errors, requiring 'quantum error corrections' (QEC) to protect qubit information and build scalable, robust systems; Kazu Nakashigi likens this challenge to the early days of classical computers in the 1980s.
Yakumo employs a neutral atoms methodology for quantum computing, which differs from Google and IBM's superconducting qubits (requiring extreme cooling) or other methods using ions or photons.
Yakumo grew to nearly 70 employees within 15 months of its April 2025 launch, with 20-30% of its tech staff being non-Japanese, including two of three division heads, leveraging Kyoto's beauty and Kyoto University's deep research background in neutral atoms.
Yakumo aims to be a 'Toyota of quantum computers,' focusing on assembling a full-stack system from specialized components like high-fidelity lasers from Hamamatsu Photonics (Japan) and Nic Photonics (Denmark), rather than building every part.
AI significantly impacts quantum computing, particularly for 'AI for quantum' applications where classical computers and AI/LLMs are used to estimate and correct errors in quantum calculations.
'Quantum for AI' is a future prospect where unique data generated by quantum computers, such as from fluid dynamics calculations, could make AI models smarter, though practical examples are not yet available.
Quantum computers are significantly less power-hungry than classical computers; for certain problems with 40-50 qubits, quantum machines can consume less energy, potentially aiding CO2 reduction.
Kazu Nakashigi believes quantum computers will function as accelerators in hybrid systems alongside classical computers, sharing workloads to make the overall computing environment smarter, rather than fully replacing existing systems.
Kazu Nakashigi suggests that RSA cryptography could be broken by quantum computers within the next 5 to 10 years, warning that encrypted data collected now could be decrypted later, posing a national security risk.
The US government aims for a 'very useful computer' by 2028, reflecting aggressive competition in quantum computing among leading nations like China, the United States, EU, and Japan, which do not fully share their scientific progress.
Karthik highlights the high current cost of quantum computing, with Toshiba's SQBM costing $90/hour and IBM's service costing $3,000/hour.
Kazu Nakashigi claims Yakumo's neutral atom approach could be 100-1000 times cheaper per qubit than superconducting qubits, which can cost up to $1 million per physical qubit.
Elon Musk is integrating Starlink into cybercabs, demonstrating cross-company technological coordination and enabling gigabit connection speeds for future vehicles. This will also facilitate distributed inference edge computing using the vehicles' onboard GPUs.
Nvidia announced record Q2 earnings, with revenue reaching $81.6 billion and EPS at $187, both surpassing estimates, propelled by a 92% growth in data center revenue.
Jensen Huang stated Nvidia's hyperscaler revenue grew 12% quarter-over-quarter, comprising 46% of total revenue and indicating market share gains, while also confirming Nvidia sells no chips in China.