NIO ended 2025 up 26%, but the journey was wild. XPeng kept over-promising. OpenAI just became too big to fail. And $160 million in Nvidia chips got smuggled to China. This is my 2025 year-end review covering everything that mattered.Thank you for rocking with Courtside Financial all year. Happy New Year! Let me know what content you want to see in 2026.NIO'S 2025 FULL STORYNIO started at $4.55, peaked at $7.89 in October, closed at $5.50 - up 26% for the year.Key moments: February Onvo L60 image sparked 10% jump. March capital raise tanked stock to $3.75, but $4B Hong Kong offering funded growth. April brought Shen Fei as new Onvo president. July-August surge: L60 pre-sales drove 40% gain, third-gen ES8 launch added another 23% spike. September $1B raise initially dropped 9% but rallied 30% as investors recognized growth funding. Stock peaked October 2 at $7.89. November declined 24% as China subsidies ended. December 30 brought 2026 subsidy news, stock closed year at $5.50.Bottom line: NIO executed three-brand strategy, launched Onvo and Firefly, improved margins, raised capital twice. Now faces proving profitability at scale.XPENG'S OVER-PROMISING PROBLEMHe Xiaopeng announced "orders exceeded expectations" five times in 2025 - only MONA M03 delivered. New P7 peaked at 8,104 September units then crashed to 2,882 by November despite 40 units/hour capacity. G7 got 10,000 orders in 9 minutes but sales fell to 2,500-3,500 units while Tesla Model Y averaged 38,000.Service failures: 7,000+ owners complained after $5,000 chip upgrades caused system crashes. XPeng promised nationwide CNGP in 2022 - still only 5 cities despite customers paying $39,000 extra. Trust matters when markets get tough.OPENAI TOO BIG TO FAIL?SoftBank invested $41 billion in OpenAI on December 26, now owns 11%. Microsoft owns 27%. OpenAI valued $300-500 billion, not yet profitable. Committed to $1.4 trillion infrastructure spending. SoftBank sold entire $5.8B Nvidia stake to fund this.The question: Has OpenAI become too big to fail? If it fails, SoftBank, Microsoft, Oracle all hurt. Data centers, Nvidia's growth, entire AI buildout slows. System has to keep OpenAI alive because letting it fail craters too much. That's "too big to fail" - saw it with banks in 2008.OpenAI burns billions, revenue growing but so are costs. If AI hype slows, they have problems. But investors can't let them fail now, will keep funding betting revenue catches up. Risk is real when you build this big, this fast on this much investment.CHIP WARS: $160M SMUGGLING BUSTDOJ unsealed Operation Gatekeeper December 8. October 2024-May 2025: Smugglers moved $160M+ in Nvidia H100/H200 GPUs to China. These chips power AI, same ones OpenAI uses. Export-controlled, can't legally sell to China.How: Fake companies, falsified documents, relabeled chips as "Sandkyan," routed through multiple warehouses. May 28 raid in New Jersey caught three trucks picking up chips. Smugglers used group chat - when police arrived, chat said "dissolve group, delete everyone." Too late.Three arrested, one pleaded guilty facing 10 years. Estimates show 10,000 to hundreds of thousands chips smuggled in 2024.Why China needs them: Building domestic chips but years behind. 60%+ of Chinese AI models still run on Nvidia. Chinese companies pay 2x-5x premium, smugglers risk prison to supply.Trump twist: Same day as bust, Trump posted US would allow H200 exports if government gets 25% cut - exact chips being smuggled. Defense attorneys asked how they're national security threats when President approves export.Reality: Export controls don't work when incentives this strong. Continues until China develops competitive chips (years away), US loosens controls (Trump willing), or AI demand crashes (not happening).

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related