Good morning everyone,
Sora, OpenAI’s new video app, is reshaping how content is made and sparking the first global debate on AI-made culture. Meanwhile, markets remain euphoric: capital is pouring in faster than revenue, and even central banks are warning the AI boom could overheat.
Sora Soars, Then Sparks a Culture War
The news of the last two weeks is the launch of Sora, a social video app that turns text into hyper-realistic clips. OpenAI’s new app became an overnight hit, topping the App Store within days. It’s the first true glimpse of what AI-made entertainment at scale looks like: fast, cheap, and endlessly remixable. Within hours, feeds filled with surreal “glurons,” cinematic vignettes, and eerily lifelike avatars.
Then came the backlash. Japan’s government accused OpenAI of training Sora on copyrighted manga and anime, calling them “irreplaceable treasures.” The Martin Luther King Jr. estate demanded takedowns of deepfake likenesses, and MrBeast warned the app could “wipe out” creator livelihoods.
OpenAI’s response was swift: CEO Sam Altman promised to license fictional characters and share revenues with rightsholders, an acknowledgment that copyright and creativity now share the same battlefield. But deeper questions remain: how do you sustain a platform where each video costs real power to generate, and where “authorship” itself is blurred? MIT Tech Review warns those issues could define AI video’s future.
A trend potentially soon impacting the rest of the creative industry as Spotify is reportedly preparing AI-generated artist partnerships. The shift is clear: creation is becoming collaboration between humans, models, and platforms, and every participant - from regulators to insurers - will soon need to decide who owns, profits from, and bears the risk of the output.
Why it matters for Zurich Strategy:
Sora’s rise isn’t just another app launch: it’s a cultural inflection point. AI content is industrializing faster than governance can follow. That means new forms of IP exposure, deepfake liability, and reputational risk, but also opportunities in emerging coverage categories.
Markets Run Hot: Trillions Flow, Warnings Grow
The AI economy is now a full-scale industrial and financial boom. OpenAI alone has locked in over $1 trillion in chip and infrastructure commitments across Nvidia, AMD, Oracle, and Broadcom—the largest capital program in computing history. The AMD deal adds 6 GW of compute from 2026, while US utilities plan $1.1 trillion in grid upgrades to meet AI-driven power demand. Together, power, chips, and data centers now anchor nearly all recent US growth.
But the exuberance is showing cracks. Both the IMF and Bank of England warn that AI-linked assets could face an “abrupt correction.” Over $1.2 trillion in AI-tied debt now dominates investment-grade markets, while ten unprofitable startups have added $1 trillion in value this year alone. The gap between spending and earnings is widening fast.
OpenAI’s $1 T+ compute deals: financed via complex equity-for-capacity swaps—raising sustainability questions as losses mount.
‘Of course it’s a bubble’: investors admit valuations are stretched but expect long-term market creation.
$1.2 T debt exposure: JPMorgan flags refinancing risk if capex slows or rates rise.
Why it matters for Zurich Strategy:
AI’s infrastructure boom is now both a growth engine and a risk multiplier. A correction in over-levered tech or utilities could ripple through credit portfolios, infrastructure funds, and power markets—tying Zurich’s investment and underwriting exposure closer to the same cycle.
97 million jobs and counting
On the workforce side, deployment is accelerating faster than policy. A US Senate report projects 97 million American jobs could be displaced within a decade as companies scale AI across operations. Deloitte alone has rolled out Claude to 470,000 employees, and Zendesk claims its new agent can resolve 80% of support issues autonomously.
But alongside productivity gains, shadow usage is rampant: 82% of enterprise AI activity happens through unapproved accounts, exposing firms to compliance and data-leak risk.
Top-performer gap widens: AI amplifies high performers’ output but leaves average workers behind, creating management and retention challenges.
Enterprise deployments scale up: Anthropic’s deal with Deloitte establishes a global “Claude Center of Excellence.”
Shadow AI usage: 45% of employees use generative AI tools; most paste data into unmanaged ChatGPT instances.
Why it matters for Zurich Strategy:
AI deployment reshapes talent pipelines, productivity metrics, and cyber/compliance exposure. Firms that manage reskilling and data governance together will outperform those treating them as separate tracks.
Quick hits
Enterprise & Adoption
Anthropic + Salesforce: Claude becomes Salesforce’s model of choice for Agentforce in regulated industries.
Microsoft + Harvard Medical School: healthcare-specific Copilot features close the gap with OpenAI.
Figma integrates Gemini AI: 13M designers now get generative visuals 50% faster.
Regulation & Policy
California regulates AI companions: new transparency and safety mandates by 2026.
Japan defends manga IP: government moves to protect cultural IP from Sora-style training data.
New York bans AI rent price-fixing: first state to outlaw algorithmic landlord collusion.
Valuations & Funding
SoftBank buys ABB Robotics for $5.4B: betting on “Physical AI.”
Reflection AI raises $2B at $8B valuation: automating software engineering with backing from Nvidia and Eric Schmidt.
Tips & Tricks
Formatting matters more than you think. A new benchmark from Improving Agents shows that the way you format data for AI models can change accuracy by nearly 20 percentage points.
The winner? A simple “key: value” structure written in plain text — far outperforming raw CSV or JSON formats. For anyone feeding data into Zurichat (whether in analysis, automation, or reporting), clean structure beats fancy syntax. Sometimes, better output starts with better input.