AI Accelerates Quantum Computing: New Studies Shake Up the Timeline for Today's Encryption

Google and Oratomic: AI-Driven Quantum Algorithm Breaks Encryption Sooner Than Expected

In April 2026, Google's Quantum AI team and the startup Oratomic published research that shakes a cornerstone of today's IT security. Their papers show that quantum computers could break standard schemes such as elliptic curve cryptography with far fewer resources than previously assumed. Google calculates that fewer than half a million physical qubits might be enough to crack keys in minutes. That is still beyond current machines, but roughly ten times less than earlier estimates.

Artificial Intelligence as a Tool for New Quantum Math

The role of AI is notable. According to the Oratomic authors, artificial intelligence was instrumental in developing the algorithm. AI efficiently searched the vast space of possible quantum circuits. This shows that AI is not only useful for writing text, but also for discovering new mathematics and hardware paths.

Cloudflare Pulls Post-Quantum Migration Forward to 2029

The consequences are concrete. Cloudflare has already announced that it is pulling forward its post-quantum cryptography deadline to 2029. Inside banks, public agencies, and energy companies, discussions are intensifying about when to migrate to standards published by the U.S. NIST, such as ML-KEM and ML-DSA. Bitcoin and other cryptocurrencies are also in the spotlight, because wallet security relies on elliptic curve math.

Reality Check: Harvest-Now-Decrypt-Later as a Real Risk

Important context: there is no cryptographically relevant quantum computer today. Current machines have thousands, not hundreds of thousands of stable qubits. But the so-called harvest-now-decrypt-later risk is real. Attackers can collect encrypted traffic today and decrypt it later, once machines are strong enough.

Why it matters

Anyone who is not thinking about a migration plan in 2026 is already behind. For organizations, that means a cryptographic inventory, a roadmap for hybrid schemes, and a review of vendor and long-term contract obligations. TIME calls the new findings a genuine shock, and warns that the world is not prepared. Long-lived data is especially exposed: patient records, contract clauses, patent information, and state secrets.

Practical example

A German machine builder today sends design data encrypted to Asian suppliers. The business value of those files spans 15 to 20 years. If an attacker records the encrypted stream today, they could decrypt it in 2032 or 2035 using a quantum computer. The answer is a hybrid setup: TLS channels that use classical encryption and ML-KEM from the NIST standard in parallel. Tools such as OpenSSL 3.5 and current Cloudflare configurations already support this. Migrating today protects data that will still matter ten years from now.