The FBI has officially blamed North Korean state-sponsored hackers for a record $6.3 billion theft from Binance, while NVIDIA has quietly revised specifications for its upcoming Rubin GPU architecture as production hardware begins shipping to hyperscalers.
Listen to today’s briefing (podcast)
Key takeaways
FBI attributes $6.3B Binance hack to North Korea’s TraderTraitor group, the largest crypto heist on record
Binance resumes phased withdrawals after covering 95% of losses, processing $4.6B in outflows on day one
NVIDIA reduces Rubin GPU memory bandwidth by 11% and NVLink by 17% while increasing scale-out networking
Rubin NVL72 systems now shipping to hyperscalers with revised 19.2TB/s HBM4 bandwidth per GPU
Binance Crypto Heist North Korean Hackers $6.3 Billion
The FBI officially attributed the theft of more than $6 billion in virtual assets to TraderTraitor, a North Korea-aligned elite hacking group also tracked as Jade Sleet, Pressure Chollima, PUKCHONG, Slow Pisces, and UNC4899, according to Siber Güvenlik Haberleri. The agency confirmed the attack occurred on or about October 29, 2026, making it the largest cryptocurrency heist on record.
Blockchain intelligence firms TRM Labs and Chainalysis had already linked the theft to TraderTraitor, which has a history targeting cryptocurrency companies including the $387.5 million Bitget theft in September 2026 and the $1.5 billion Bybit heist in February 2025. Attackers stole approximately $6.36 billion across ten blockchains, with most funds transferred in just 20 minutes, primarily converting assets to Ethereum for further laundering through DeFi platforms.
Binance linked the breach to a compromised DevOps engineer’s laptop that allowed unauthorized withdrawal transactions, and stated it was preparing to resume withdrawals on November 1 between 06:00 and 09:00 UTC. The exchange said it had covered approximately 95% of the impacted amount, with BTC, ETH, USDT, BNB, SOL, USDC, XRP, USD1, RLUSD, and ZEC balances fully backed, and committed to bearing the financial impact without passing losses to users.
Why it matters: This heist underscores the critical need to secure DevOps pipelines, monitor for social engineering like Operation DreamJob, and implement rapid incident response for blockchain-based attacks.
NVIDIA Rubin GPU Specifications Updated Production Hardware
NVIDIA has quietly revised Rubin GPU specifications as production hardware ships to hyperscalers, reducing HBM4 memory bandwidth from 22TB/s to 19.2TB/s per GPU (an 11% decrease) and NVLink interconnect bandwidth from 3.6TB/s to 3.0TB/s (a 17% reduction), according to servethehome.com. The updated specs were posted in September on NVIDIA webpages and in its architecture datasheet, initially spotted by IT Home.
Scale-out networking bandwidth increased from 400GB/s to 450GB/s per Rubin GPU, while theoretical compute throughput remains unchanged at 50 PFLOPS FP4 inference and 35 PFLOPS FP4 training. NVL72 rack cumulative GPU memory bandwidth was revised from 1,580TB/s to 1,400TB/s, with NVLink bandwidth across the rack dropping from 260TB/s to 216TB/s, while scale-out networking rose from 28.8TB/s to 32.4TB/s bidirectional.
Other Rubin-based products also received revisions including the Vera Rubin Superchip, Vera Rubin NVL4, and HGX Rubin NVL8. NVIDIA previously noted specifications were preliminary and that all values were subject to change. AMD’s competing MI455X runs HBM4 stacks at 7.6Gbps/pin compared to NVIDIA’s revised 9.375Gbps/pin, which remains above the official HBM4 spec of 8Gbps/pin.
Why it matters: The revised specs affect AI workload performance planning, particularly memory-bandwidth-sensitive tasks and scale-up versus scale-out architecture decisions.
Hardware Wallet on Amazon — Secure your crypto assets after major exchange hack. See current options and reviews on Amazon. View on Amazon
As an Amazon Associate I earn from qualifying purchases.
Both developments highlight how state-sponsored threats and hardware specification changes continue to shape security and infrastructure strategies for technology professionals.
Ali YAZICI is a Senior IT Infrastructure Manager with 15+ years of enterprise experience. While a recognized expert in datacenter architecture, multi-cloud environments, storage, and advanced data protection and Commvault automation , his current focus is on next-generation datacenter technologies, including NVIDIA GPU architecture, high-performance server virtualization, and implementing AI-driven tools. He shares his practical, hands-on experience and combination of his personal field notes and “Expert-Driven AI.” he use AI tools as an assistant to structure drafts, which he then heavily edit, fact-check, and infuse with my own practical experience, original screenshots , and “in-the-trenches” insights that only a human expert can provide.
