Samwise Tech/AI/Robotics Newsletter
Monday, July 13, 2026
AI Faces $3 Trillion Revenue Test as Infrastructure Spending Hits $1.5 Trillion in 2026
The AI industry must generate $3 trillion in annual revenue to justify the $1.5 trillion projected to be spent building AI infrastructure in 2026 alone, according to Sequoia Capital partner David Cahn. Cahn argues the math is unavoidable: compute infrastructure must ultimately pay for itself. Apollo Global Management chief economist Torsten Slok warns that a sustained shortfall in hyperscaler cash flows — driven by Microsoft, Google, Amazon, and Meta — could trigger a broader macroeconomic recession. The analysis comes as data center construction, GPU procurement, and large-scale power capacity deals continue accelerating worldwide with revenue returns still lagging behind investment.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Apple Sues OpenAI Alleging Trade Secret Theft Tied to Hardware Push
Apple filed suit against OpenAI in U.S. District Court, Northern District of California, alleging trade secret theft connected to OpenAI’s hardware ambitions. The complaint names Tang Tan, OpenAI’s Chief Hardware Officer and a 24-year Apple veteran, accusing him of coaching departing employees to evade Apple security measures and sharing confidential product codenames during recruiting. A second defendant, Chang Liu, a former Apple senior systems electrical engineer, allegedly downloaded confidential documents and never returned an Apple laptop. Apple sent a formal letter in February 2026 with no response before filing suit. OpenAI is building its first consumer device following a $6.5 billion acquisition of Jony Ive’s io design firm.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Nvidia Shares Down 15% Since May as DRAM Displaces GPUs as the AI Bottleneck
Nvidia shares have fallen roughly 15% since May as AI infrastructure capital shifts from GPUs to memory chips. DRAM prices have risen approximately 10 times since summer 2025, while Micron Technology shares have nearly tripled over the same period. The trend reflects a market view that the acute GPU shortage has eased, replaced by memory bandwidth as the binding constraint in large AI training and inference clusters. H100 spot instance prices peaked near $3.20 per hour before declining. Wayne Nelms, CTO of cloud-infrastructure analytics firm Ornn, is among industry observers tracking the shift. Meta, Google, and Amazon are accelerating custom silicon development to reduce merchant GPU dependence.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Meta’s Custom MTIA AI Chip Enters Mass Production at TSMC in September
Meta will begin mass production of its custom MTIA AI inference chip at TSMC in September, after completing approximately six weeks of testing, the company announced. The chip was co-designed with Broadcom; Samsung supplies DRAM and Sandisk provides storage. The MTIA rollout is central to Meta’s plan to deploy 7 gigawatts of AI compute capacity, funded by a capital expenditure budget of $125 billion to $145 billion this year. Meta joins Microsoft, Google, and Amazon in accelerating custom silicon development to bring inference costs in-house and reduce dependence on Nvidia GPUs as AI workloads continue to scale.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Robot.com Launches R-noid Humanoid for High-Turnover Commercial Settings
Robot.com has launched the R-noid, a humanoid robot aimed at repetitive commercial tasks in sectors with acute turnover problems. Quick-service restaurants, one primary target, average 130% annual staff attrition. The company offers the robot under a robotics-as-a-service subscription model, with deployments taking 8 to 12 weeks from site assessment to full operation. The R-noid addresses five task categories across six verticals, covering 19 defined workflows. The robot runs on the π0.7 motion model from Physical Intelligence and uses an Nvidia Jetson platform with Nvidia Isaac Sim for simulation and training. Robot.com positioned the launch around addressing burnout-prone roles rather than broad workforce displacement.
Sources: Robotics & Automation News Share ↗ ✉︎ Email 💬 Text
FORT Robotics Joins Nvidia Halos Ecosystem to Bring External Safety Layer to Physical AI
FORT Robotics has joined the Nvidia Halos for Robotics ecosystem, integrating its Outside-In Safety Blueprint with Nvidia’s physical AI safety platform. Unlike onboard safety systems, FORT’s approach uses external infrastructure sensors and visual AI agents to monitor robotic environments in real time. The integration runs on Nvidia IGX Thor hardware with Holoscan Sensor Bridge. Nvidia simultaneously announced the world’s first ANAB-accredited inspection laboratory for physical AI systems — the Nvidia Halos AI Systems Inspection Lab — with FORT among the first qualified partners. CEO Samuel Reeves described safety as “the precondition for scale” in physical AI deployments, framing the external safety layer as essential to broad commercial robotics adoption.
Sources: Robotics & Automation News Share ↗ ✉︎ Email 💬 Text
STMicroelectronics Takes Stake in Oversonic Robotics, Maker of Italy’s First Certified Cognitive Humanoid
STMicroelectronics has acquired a minority stake in Oversonic Robotics, an Italian company that developed RoBee, described as the world’s first certified cognitive humanoid robot. The investment round also includes Fondazione ENEA Tech Biomedical, Italy’s national energy and technology research foundation, and SpotInvest. Oversonic plans to use the capital to expand into the United States. STMicroelectronics, a major European semiconductor manufacturer with broad embedded systems and IoT product lines, enters the humanoid space through investment rather than direct development. The deal reflects a trend of semiconductor companies positioning within physical AI supply chains as humanoid robot commercialization accelerates globally through 2026.
Sources: Robotics & Automation News Share ↗ ✉︎ Email 💬 Text
Tech Pulse
Top Frontier Models (SWE-bench Pro): Claude Mythos 5 (80.3%) | Claude Fable 5 (80.0%) | Sakana Fugu-Ultra (73.7%)
Top Open Source Models (MMLU): Qwen3.5-27B (91.5%) | DeepSeek R1-0528 (90.5%) | Qwen3-235B-A22B (88.7%)
Top Small Models (15–50B, MMLU): Gemma 4 31B (89.0%) | Qwen3 30B-A3B (85.3%) | Mistral Small 3.2 24B (76.0%)
Top Edge Models (0–15B, MMLU): Phi-4 14B (77.6%) | Llama 3.1 8B (63.8%) | Gemma 3n 4B (62.9%)
AI Leaders: Nvidia $4.75T | Apple $4.59T | Alphabet $4.31T
Robotics Leaders: Intuitive Surgical $175.2B | Teradyne $42B | Fanuc $40B
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.