TechFlow Intelligence: Bitcoin Drops Below $80,000 Amid AI Sell-Off Wave, SpaceX Raises $40 Billion to Buy GPUs
AI / Large Language Models
DeepSeek Completes Approximately 100 Billion Yuan in New Financing, Shattering China's AI Industry Single-Round Record
Generating only hundreds of millions in revenue while raising 100 billion yuan, the valuation logic bets entirely on future computing power and generational model gaps. Synchronous data from CAICT shows daily domestic token consumption averages around 500 trillion, and large model token prices have doubled compared to the second half of last year.
> Hot Take: Raising 100 billion on hundreds of millions in revenue isn't financing; it's a casino pre-selling the future.
Source
OpenAI and Anthropic Slash Large Model API Prices by Up to 80% Within Months
The price cuts from these two leading labs far exceed any previous round, directly squeezing the survival space of mid-to-small inference service providers. On Hacker News, 333 comments are debating whether this constitutes a "dumping-style market clearance."
Source
Musk Confirms SpaceX Secures $40 Billion Debt Financing for GPU Procurement
SpaceX secured this funding using investment-grade credit, with capital expected to arrive in Q1 and Q2 of 2027. The cash flow from Starlink and rocket launches has become collateral for the AI computing arms race.
> Hot Take: A rocket company borrowing money to buy GPUs means the endgame of the AI bubble is a space data center.
Source
Inference Service Providers Are Erupting in Number, Earning Millions a Month in Just Weeks
Open-source models like GLM, Kimi, and DeepSeek are running inference on rented B300 cards, achieving gross margins of $10-20 per GPU per hour. However, Chinese model labs are beginning to demand revenue sharing, pushing entry barriers higher again.
Source
OpenAI Bans Direct Internet Access During State-of-the-Art Model Training
The training environment is isolated from the public internet, ostensibly to prevent data contamination and adversarial injection. The community speculates whether this is a new security protocol or a fear that models will "learn" undesirable things.
Source
Moonshot AI Launches Kimi K3.1 Pro, High-Speed Version Booms 2-3x Faster
Around the same time, Wuwen and Three-Body Universe jointly released Sophon Agent 1.0, with domestic models ramping up investments in both speed and IP commercialization.
Source
Crypto / Web3
Bitcoin Dips Below $80,000, Cited by Bloomberg as Result of Deepening AI-Related Market Selloff
The AI computing narrative and crypto markets retraced in sync, strengthening correlations between risk assets. An Ethereum Foundation-linked address also sold 2,785 ETH at this level.
> Hot Take: The foundation always sells at the peak. This time, the broader market dipped first out of respect, and they're responsible for the "blatant finishing blow."
Source | Source
Bitcoin Wallet Vulnerability: Malicious Servers Can Steal Funds When Users Speed Up Payments
The issue lies in the fee replacement process, where attackers can craft malicious responses to trick user wallets into signing unintended transactions. A PoC has been publicly released on GitHub.
Source
Solana Users Holding Tokenized Stocks Reach 1.5 Million, Setting an All-Time High
This figure represents a fivefold increase from August, as on-chain stocks transition from geeks' toys to a genuine user base.
Source
Franklin Templeton to Launch "US Stocks + Bitcoin" ETF, Automatically Buying BTC with Stock Dividends
Targeting a 5% BTC allocation, the essence is allowing ordinary investors to passively gain Bitcoin exposure while buying stocks.
Source
Entity Holding Aave, Worldcoin, and ETH Transfers ~$45.38 Million in Tokens to Multiple Exchanges Within Two Days
Including 35.1 million WLD, 6,000 ETH, and 97,000 AAVE, on-chain data clearly points to a large-scale exit maneuver.
Source
Chips / Hardware
NVIDIA Officially Lowers HBM4 Pin Speed from 11.7Gbps to 9.6Gbps
Simultaneously, NVLink bandwidth per GPU drops from 1.8TB/s to 1.5TB/s. AMD is now actually leading in both HBM bandwidth and scale-up bandwidth metrics.
> Hot Take: Jensen Huang voluntarily lowering speeds isn't a change of heart; it's because HBM4 yields and supply simply can't hold up.
Source
Memory Chip Shortage to Persist Until 2028, Nomura Says Sector Valuation Is Severely Underpriced
Korea's "memory duopoly" is expected to report over 184 trillion KRW in operating profit for Q3, more than double the same period last year. Memory price hikes have already caused global PC shipments to plummet 21% in Q3.
Source
Jensen Huang: Open to NVIDIA Collaborating with CXMT
The remarks also included "no country is stronger than China in manufacturing," sending a clear supply chain signal amid memory shortages.
Source
Huaqiangbei GPU Prices Fluctuate Daily, Rumors of NVIDIA Production Cuts Shake the Market
Overlapping computing demand and supply panic have driven retail price volatility to levels approaching those of the mining boom era.
Source
Tech Companies
iPhone Duo Sells Out Immediately Upon Launch, Apple Demands 30% Supply Chain Production Increase
All models sold out on day one, with the next shipment scheduled for January 22, 2027. However, teardowns by iFixit and analyses by YouTubers all point to inner screen damage at the same specific location.
Source | Source
IBM CEO: Open-Weight Models Locally Deployed Are Now "Too Good to Resist"
Emphasizing "good and cheap," this marks the most explicit endorsement yet by a traditional enterprise IT giant toward open-weight models.
Source
Finance / Macro
Gold Breaks Below $4,000, Markets Bet on Fed Rate Hike in December
US inflation consistently beats expectations, while employment and manufacturing remain resilient, reigniting Fed tightening expectations. As a "non-yielding asset," gold faces sustained selling pressure in this environment.
Source
Oil Prices Hit Five-Month High, US-Iran Conflict Risks Drive Up Global Risk Premium
Brent crude briefly neared $115.6, with markets simultaneously fearing disruptions in the Strait of Hormuz. The Houthis claimed attacks on Saudi Arabia's east-west oil pipeline terminals.
Source
US Tech Giants Face Earnings Tests: Azure Up 43%, Google Cloud Up 82%, AWS Up 37%
All cloud revenues show strong growth, but massive AI infrastructure investments lead markets to question "whether growth can defend valuations." Interest rate decisions from the Fed, ECB, and BOJ will all be released this week.
Source
New Products / Trends
One Day After iPhone Duo Launch, Massive Online Reports of Identically Located Inner Screen Damage
On Zhihu, 539 comments debate whether this is a design flaw or user misuse. "Crackgate" has already flooded English-speaking communities.
Source
Unitree Robotics Share Price Briefly Hits 700 Yuan, Humanoid Robot Concept Remains Heated
The A-share Shanghai Composite Index opened low and closed high, up 0.19%, with market sentiment primarily driven by the robotics sector.
Source
Today's Hidden Thread
Everything today tells the same story: AI's physical constraints are beginning to bite back at financial narratives. Bitcoin breaking below $80k is attributed to an "AI selloff wave," NVIDIA quietly lowering HBM4 speeds exposes memory supply chain bottlenecks, yet SpaceX is still borrowing $40 billion to buy GPUs—the scarcity and pricing signals of computing power are rippling from chips to energy, and from US equities to crypto markets.
Meanwhile, the US-Iran conflict pushed oil prices to a five-month high, adding another layer of inflationary pressure to this already strained computing narrative. Gold dropping below $4,000 isn't due to vanishing safe-haven demand, but rather a sudden market realization: rates may truly need to rise further. When the AI arms race collides with high interest rates and high oil prices, all "cost-is-no-object" narratives must face a reckoning.