Sleepless AI (AI) Surge

Amidst the AI investment boom on Wall Street, a rare bond issuance frenzy is sweeping major US tech firms. AMD recently raised $4.75 billion in its largest-ever USD bond sale. This follows massive debt offerings from Nvidia ($25B), Amazon (planning at least $25B), and Alphabet ($25B), with tech giants collectively issuing nearly $220 billion in bonds so far this year—more than double 2025's total. While this fuels an AI infrastructure "race," raising real yields, a hidden risk looms. Research by Goldman Sachs and Morgan Stanley reveals nearly $2 trillion in off-balance-sheet financial commitments from giants like Alphabet, Microsoft, Amazon, and Meta. These stem from future lease obligations and purchase contracts for data centers, not formally recorded as debt. The trend, pioneered by Meta's structured financing for its Louisiana data center, allows massive expansion without bloating balance sheets. Though traditional credit metrics for these companies remain strong, credit markets are reassessing risk. Yields on related bonds are rising, and credit default swap (CDS) premiums for firms like Oracle and Nvidia have hit record highs. The concern is that if AI data center returns disappoint, these hidden liabilities could significantly impact credit markets. Investors are advised to look beyond headline numbers and scrutinize financial statement footnotes.

Manus, an AI company that experienced a meteoric rise in March 2025, a $2-3 billion acquisition by Meta, a regulatory block in China in April 2026, and an eventual buyback by its original Chinese investors, has regained its independence. Its relaunch coincides with the formal implementation of China's new "Regulations on Outbound Investment" (State Council Decree No. 837) as of July 1, which emphasizes compliance from a venture's foundational structure. Analysis suggests Manus became a cautionary tale due to "opportunistic restructuring"—moving its main entity to Singapore during a policy gray area after gaining significant user traction, rather than establishing a compliant global architecture from inception. Its core competency was seen as product integration and first-mover advantage, not insurmountable technical barriers, leaving it vulnerable when regulatory and scrutiny pressures mounted. The incident, however, hasn't cooled investor enthusiasm for AI出海 (AI going global). Capital remains active but more discerning, now heavily prioritizing ventures with legally sound, forward-designed global structures over reactive "re-domiciling." A contrasting example is Singapore-based ccMonet.AI, which established its commercial and legal headquarters overseas from day one, outsourcing R&D to China for efficiency while maintaining clear compliance. A key trend emerging is the growing bifurcation for AI startups: deeply serving either the domestic Chinese market or focusing squarely overseas, as straddling both becomes increasingly difficult. For those targeting global markets, a clear strategic choice is needed between keeping core R&D in China (avoiding tech export issues) and pursuing overseas sales, or fully integrating R&D and sales abroad under a specific national jurisdiction, each path carrying distinct regulatory implications. Furthermore, the choice between To B and To C models is critical. While Chinese founders are seen as highly competitive in global consumer apps, breaking into enterprise sales, especially in markets like the US, presents significant Go-to-Market challenges requiring localized sales teams once beyond a certain scale. In conclusion, the new regulations and Manus's saga highlight that "architecture-first" compliance is no longer optional but essential for survival in AI出海. The era of exploiting regulatory gray areas carries high deferred costs, as Manus's dramatic year demonstrates. Ultimately, while solid legal architecture determines if a company can survive, product strength and profitability determine how well it thrives.

Cashi has launched a payment card in Hong Kong that allows users to spend stablecoins and earn cashback, with Thredd serving as the AI-powered payment processing partner. The Cashi app enables users to store, transfer, and spend digital assets for everyday expenses like groceries, subscriptions, online shopping, and travel. Currently, virtual Visa cards and Google Pay are available in Hong Kong. The company plans to expand to Mexico by the end of 2026, where it will introduce physical cards and Apple Pay support. This launch in Hong Kong and Mexico is intended as a stepping stone for Cashi's further global market expansion.

OpenAI has launched a new feature called "Computer History" for its macOS desktop ChatGPT application. Designed for Pro, Business, and Enterprise subscribers, the tool monitors user activity across applications and websites in the background. This data builds a temporary activity timeline, allowing the AI to reference it in subsequent conversations to provide more accurate and personalized responses while reducing the need for manual context input. The feature, an evolution of the earlier "Chronicle" project, requires explicit user activation in settings and offers privacy controls, including the ability to clear all or selected portions of the history. OpenAI emphasizes its operation is based on voluntary consent. However, the rollout is delayed for users in the European Economic Area, UK, and Switzerland as OpenAI works to align the feature with local data protection regulations like GDPR. This highlights the ongoing challenge of balancing advanced personalization with strict data privacy laws. The announcement coincides with other technical advances from OpenAI, such as the faster GPT-5.6 Ultrafast API model. The article's analysis notes that while this move intensifies competition with rivals like Claude and Gemini, it also raises questions about long-term user trust, potential security risks present in earlier versions, and whether voluntary consent remains a sufficient safeguard as the scope of collected activity data expands.

The article discusses the shift of investment capital from cryptocurrencies to artificial intelligence (AI) sectors, as confirmed by data showing AI captured 80% of global venture funding in Q1 2026. This has contributed to stable but range-bound cryptocurrency prices due to reduced market interest and liquidity. While crypto analyst Arthur Hayes asserts that AI has absorbed all available capital, Jeff Park, a partner at ParaFi Capital, presents a contrarian view. He argues that the wealth creation from AI is ultimately beneficial for Bitcoin. Park believes that once the inevitable AI bubble bursts and capital flees, Bitcoin, as a perpetual asset, will be a primary destination for that liquidity. This prediction has drawn criticism from those who warn of a potential correlated downturn between stock and crypto markets if the AI sector collapses.