What technical markers truly distinguish 2024 from previous years in the high-tech sector? Three main areas stand out from the announcements: the shift of AI computing to the devices themselves, the publication of post-quantum cryptography standards by NIST, and the entry into force of the European AI Act. Rather than a catalog of new features, this article measures the gap between these advancements and what existed before them.
Post-quantum cryptography: NIST standards published in 2024
Competitors widely cover AI and 5G, but overlook a technical milestone published in August 2024: NIST has finalized three post-quantum cryptography standards. These algorithms are designed to withstand attacks that future quantum computers could launch against current encryption systems.
The issue is not theoretical. Data encrypted today with classical protocols (RSA, ECC) can be intercepted and stored right now, then decrypted later by a sufficiently powerful quantum machine. This scenario, often referred to as “harvest now, decrypt later,” primarily concerns sectors that manage long-lived data: health, defense, finance, administration.
To keep up with the news of these changes, Cyber Huge’s high-tech articles regularly cover developments related to digital security and emerging technologies.
The migration to these new standards involves a complete inventory of an organization’s cryptographic assets: certificates, software libraries, communication protocols. Migrating is not just changing a password, but auditing every layer of the information system. It is an infrastructure project, not just a simple software patch.
On-device AI vs. cloud AI: 2024 approach comparison table
Processing artificial intelligence directly on the device (on-device) marked a turning point in 2024. Qualcomm, MediaTek, Google with Gemini Nano on the Pixel 8 Pro, and Apple with Apple Intelligence have all invested in neural processing units (NPUs) integrated into chips. The table below compares the two architectures.
| Criterion | Cloud AI (pre-2024) | On-device AI (2024) |
|---|---|---|
| Latency | Depends on network connection | Almost instantaneous response on the device |
| Privacy | Data transmitted to a remote server | Data processed locally, without transit |
| User-side energy consumption | Low (remote computation) | Higher locally, but eliminates network cost |
| Offline operation | Impossible or severely degraded | Operational without network |
| Complexity of executable models | Large models, few constraints | Compressed models, adapted to limited resources |
The clearest gap concerns privacy. A device that translates in real-time, generates summaries, or personalizes an interface without sending data to a server profoundly changes the relationship between the user and the service. On-device AI repositions privacy as a hardware argument, rather than just a simple privacy policy.
Energy footprint of AI data centers
The shift to on-device does not eliminate the overall energy consumption of AI. The International Energy Agency identifies artificial intelligence as a driver of the expected increase in electricity demand from data centers. AI reduces latency for the end user but increases the load on training infrastructures.
Embedded models must first be trained on server farms before being compressed to fit on a mobile chip. Therefore, the gain in user-side privacy coexists with a heavy energy bill on the developer side.
European AI Act: a phased implementation timeline
The AI Act came into force in 2024, but its obligations do not apply all at once. The regulation distinguishes several categories of systems and staggers the deadlines:
- AI practices considered prohibited (generalized social scoring, subliminal manipulation) are the first targeted by the timeline.
- General-purpose AI models, such as large language models, are subject to specific transparency and technical documentation obligations.
- High-risk systems (automated recruitment, medical devices, biometric identification) will have to comply with enhanced requirements according to a separate timeline.
This phased approach changes how digital companies plan their compliance. The AI Act imposes continuous technical auditing, not a one-time certification. A company using a language model for customer support does not face the same deadlines as a medical imaging software publisher.
Mandatory USB-C: a complementary regulatory signal
The European Union also mandated in 2024 that USB-C charging be the standard for electronic devices, accompanied by rules on consumer information at the point of sale. This type of hardware standardization aligns with the logic of the AI Act: the European regulator now intervenes as much on software as on the physical connector.
Tech trends 2024: what the data reveals about the direction of the sector
The three areas analyzed here, post-quantum cryptography, on-device AI, and European regulation, share a common trait: they shift technical responsibility upstream. Securing data against a future quantum threat, processing AI locally rather than in the cloud, complying with a phased regulation: each case requires structural anticipation, not a reaction to a product launch.
The dominant trend of 2024 is not a gadget, but an architectural change. NPU chips, NIST standards, and the AI Act are reshaping the foundations on which the next generations of digital products and services will be built. Organizations that audit their cryptographic systems and AI processing chain now are positioning themselves on ground that regulation will soon make mandatory.