The year 2024 marks a turning point for several technological bricks that are moving out of the experimental stage and into production. Three axes structure this evolution: the shift of artificial intelligence to devices, the tightening of European regulations, and the reorientation of investment flows towards deep tech. Here, we analyze the concrete implications for IT architectures, data strategies, and partner choices.
On-device AI and Edge AI: the architectural break that redistributes value
The most underestimated movement of 2024 does not concern the language models themselves, but where they run. Qualcomm, Apple (with Apple Intelligence), and Google are pushing AI models running directly on smartphones and PCs. The reasons are threefold: reduced cloud inference costs, native user data privacy, and near-zero latency.
See also : Discover the latest quirky and entertaining web trends in 2024
This on-device shift redraws the value chain. Where a classic generative AI service charges for each API request, an embedded model shifts the center of gravity to the chip and firmware. For IT departments, this means rethinking supplier trade-offs: the performance of a device’s NPU (Neural Processing Unit) becomes a purchasing criterion on par with RAM or storage.
We are already observing consequences on hybrid architectures. Sensitive processing (internal document analysis, voice transcription, contextual suggestions) migrates to the device, while heavy tasks (training, long generation) remain in the cloud. This distribution requires a review of data pipelines and synchronization policies, a project that most companies have not yet budgeted for.
Related reading : The Latest Real Estate Market Trends to Watch in 2024
Analyses published on the neonews.fr site dedicated to tech regularly detail the benchmarks of these NPUs and their impacts on professional device fleets.
AI Act, DMA, DSA: what European regulation changes in tech roadmaps
The European AI Act enters its phase of gradual implementation in 2024. High-risk AI systems and foundation models will have to meet obligations for transparency, data governance, and risk management. This is no longer a weak signal: it is a compliance timeline with firm deadlines.
In parallel, the Digital Markets Act (DMA) and the Digital Services Act (DSA) are producing their first concrete effects. Major platforms must open their ecosystems (interoperability of messaging services, access to advertising data) and strengthen their moderation mechanisms. For marketing departments and data teams, advertising strategies based on walled gardens are being challenged.
Direct implications for technical choices
- AI models deployed in Europe now require complete technical documentation (training datasets, identified biases, performance metrics), which lengthens production cycles.
- The interoperability mandated by the DMA forces a rethink of integrations with dominant platforms, particularly for customer communication and user data management.
- The moderation obligations of the DSA increase compliance costs for any company operating a digital service with user-generated content.
The key takeaway: regulation does not hinder technological adoption; it changes the sequence. Companies that integrate compliance from the design phase (privacy by design, continuous documentation) save time compared to those that treat regulation as an added layer at the end of the project.
Deep tech and reorientation of investments: the end of generalist funding
The venture capital market in 2024 marks a clear shift. Investments in generalist tech startups are declining in favor of deep tech, which refers to companies built on fundamental scientific advancements: specialized chips, advanced materials, computational biotechnology, quantum systems.
This repositioning has a technical explanation. Application layers (SaaS, marketplaces, aggregators) are reaching a ceiling of differentiation. Investors are seeking sustainable barriers to entry, and these barriers are now found in hardware and fundamental research.
What this means for user companies
Technological partnerships looking 3-5 years ahead must integrate the deep tech component. A cloud provider that does not develop its own inference chips (as Google does with its TPUs) risks losing competitiveness against those who master the complete stack.
For industrial sectors (energy, health, defense), deep tech opens direct applications: more precise sensors, faster simulations, materials with programmable properties. The decision cycle is longer, but the competitive lock-in is much stronger than a software advantage.
Generative AI in business: beyond the prototype, the production constraints
The majority of large companies have launched generative AI pilots. The challenge of 2024 is no longer experimentation; it is industrialization. Gartner anticipates that the share of companies using GenAI in production will grow massively by 2026, which implies solving three concrete problems.
- Management of trust and AI-related risks (AI TRiSM): monitoring models in production, detecting drifts, protecting training data. Without these safeguards, the benefits of generative AI are nullified by reputational and legal risks.
- The cost of large-scale inference: each request to a foundation model has a unit cost. Multiplied by millions of customer interactions, the budget can exceed that of traditional infrastructure.
- Integration with existing systems: connecting a LLM to an ERP or CRM is not just about an API. It requires structuring internal data, managing access rights, and ensuring the freshness of information injected into the model’s context.
Companies that move from pilot to production in 2024 are those that have addressed these three points from the scoping phase, not after the first incident.
The technological landscape of 2024 is characterized by a shift in the center of gravity: from applications to hardware, from centralized cloud to edge, from generalist funding to fundamental research. Technical departments that adjust their roadmaps along these three axes will gain a measurable lead in the following cycles.