Tech in 2026: A JP Global OpEd

Looking back to 2025, AI still dominated the technology landscape as it did in 2024 when Agentic AI was touted to be the next evolution of AI capabilities with its promises of automating everything. While agents infiltrated business as ‘digital colleagues’ it’s fair to say they didn’t live up to the hype. The positives are obvious when starting from zero on a new tech as the only way is up but as those that had already tried to scale a RAG system without losing performance or to those that understood the futility of trying to use an inherently non-deterministic LLM architecture to perform deterministic tasks, the experts knew the negatives. The biggest win of the year for AI of the year was democratising technology, novices could use AI to masquerade as experts. It is an illusion though without experience but AI masks this well with its core strength of being able to synthesise information into a coherent and assured response. Until it doesn’t.

AI Vibes

The AI buzzword that hit the zeitgeist last year was undoubtedly 'vibe coding'. For better or worse the number of non-developers developing apps rose dramatically and the developers who were already developing apps are significantly more productive. Ignoring the drastic reduction in code quality by almost every metric, this democratisation of development will have a long term positive impact but in 2026 we will have to ride out the wave of mediocre products that are flooding the market, while keeping a keen eye on the ones that rise to the top. While vibe coding has democratised development, it has ignored the engineering required to do it successfully. In 2026, the ‘vibe’ will meet reality, forcing a return to the fundamentals of requirements specifications, system architectures, data engineering, test frameworks, operational processes, technical debt etc. Despite this 32% of global businesses expect a decrease in workforce due to the impact of AI in 2026, with only 13% expecting an increase. If you are worried about AI taking over jobs and changing the economic dynamics of the global workforce then the positive I can give you is take a look at all the vibe-coded AI slop around. While this method is extremely good for building innovative PoC’s and releasing them to Prod in a matter of hours, there’s a reason why software engineering exists and it hasn’t been idle in the past few decades. For some reason a significant number of people who should know better are forgetting or ignoring all the good practices that engineering and business have learnt the hard way already.

AI Matures

The impact of this in 2026 and beyond will be a shift in power from established technology brands to newer alternatives. Technology moves at a fast pace by default and today’s shiny new tech becomes legacy, replaced by the new kid on the block. Repeat ad infinitum. Every once in a while though this technology evolution becomes a revolution that forces this transition across a variety of products and services across industries. Cloud-enabled and cloud-native products and services were the last such technology shift, and in 2026 AI-enabled and AI-native products and services will force that transition even further than it has so far. The current state-of-the-art is currently driven forwards in a way that not even the internet could match in terms of speed of innovation and uptake. All of the technology used in a B2B and B2C market you know and love are implementing AI in some way internally or externally and that is not stopping, though it might be slowing in its current phase as the industry awaits AI engineering to develop the experience needed to truly impact outcomes. The sheer quantity and quality of AI research and development is creating a momentum that we at JP Global are specialists in leveraging with our capabilities by contributing our development of hybrid Neurosymbolic and Generative AI models and our development of Diffusion-based Real-time Frame Analysis.

The Hype Bubble

As we head into 2026 the AI hype is meeting reality like an unstoppable object meeting an immovable wall. The talk of an AI bubble grew louder in 2025 and finally started to get traction as it became obvious to even more people. By the end of the year, even the Bank of England issued a warning.  AI has transcended the boundaries of technology and its impact on each of our personal, professional and social lives could be more profound than the internet was, simply due to the ease of access. When the internet started you needed access to expensive equipment and that equipment went down in cost and increased in availability in the course of years and decades. We are still in the Web 2.0 phase with Web 3.0 waiting patiently on the sidelines. No such restraint with AI though and in 2026 the impact of AI will start to become real. Literally, as CES 2026 already proved robots are going to be everywhere. Humanoids, service robots, warehouse robots, retail robots, hospitality robots, it’s robots all the way down! The market for this new generation of AI-native robots is projected to grow at a CAGR of 40%+ through 2030, hitting a valuation of $33B by the end of the decade. Vision-Language-Action models follow natural language rather than explicit automation code and this flexibility aligned with existing human and superhuman levels of dexterity means that in 2026 AI is going to break out of the digital world and step into ours. The AI equivalent of “One small step for mankind…” as it encounters this literal new world. All the real world modelling in digital form does not hold up to actual physics though. The Achilles heel that will need to be solved though is that the training data for these advanced AI robots does not simulate actual physics. This is where companies that have actual hardware are finding that the software-based training is simply not good enough. Talking of which, agentic AI will need to mature into discrete modular low-latency embedded models able to run fully autonomously. Some way to go there so no need to worry about AI taking over. Yet.

Sledgehammers or Scalpels

The other elephant in the room for 2025 was the lack of hardware available at a reasonable cost to actually develop AI at scale. Unfortunately, none of these supply chain problems are going to disappear in 2026. The rise of AI will still be hampered by the sharp rises in RAM and GPU hardware availability and the knock-on impact this has. OpenAI, Gemini, Anthropic et al will continue to release new versions of their frontier models and new ways of interacting with them. AI will continue to creep into existing products and services, whether it makes sense or not, and undoubtedly more tech jobs will be lost as the hype bubble continues to grow, albeit at a slower rate than 2025. The stark fact remains though that less than 10% of AI initiatives actually have any sort of positive ROI. Most implementations (approx. 60%) stay in ‘experimental’ or ‘pilot’ phase and only 30% actually make it to the point of scaling towards a large-scale deployment. This is exactly what we predicted would happen in 2025 and we don’t see the lessons learnt being applied yet, so expect the trend to continue in 2026. The advance of AI will continue but the cracks in the foundations will be more apparent and the chance of the AI hype bubble popping in 2026 is obviously higher. The insights we are advising our clients and partners to navigate this in 2026 is to focus on services over products. So far the industry has favoured focussing on AI products but an analysis of what is getting traction and being successful is the AI use-cases that focus on providing a service, enhanced or enabled in a novel or innovative way by AI that wasn’t possible without the unique advantages that it brings. The truth is the rest are not very useful at best. AI technology is shoehorned where it doesn’t belong or it’s simply implemented badly, the saving grace being that the general public and end-users are still so enamoured with AI that they are willing to ignore the glaring problems. Stakeholders though who have not seen the promises delivered so far are rightly sceptical and the engineering community tasked with solving this are learning the hard way how hard AI is to deploy at scale. What did people expect when they tried to use a sledgehammer when a scalpel was needed?

Shifting Landscape

In 2026, SaaS AI services are growing at an unprecedented rate, largely driven by the abstraction and ease of being built on SaaS platforms by other SaaS services. The barrier to entry is the lowest it has ever been, hooking users in at scale faster than the infrastructure can cope. This is a double-edged sword though as costs on SaaS services scale dynamically too. For enterprises this was always a risk-balanced decision between CapEx and OpEx and now bedroom coders and weekend hobbyists will also need to think about infrastructure costs, AI is hungry and expensive at scale.

We also can’t talk about AI in 2026 without addressing the raw data, the fuel that AI feeds off. AI generating training data for AI creates a circular loop that spirals downwards in a cycle of decreasing accuracy. The promise of abundant token windows has just made everyone lazy in optimising their tokens but as I mentioned in my recent Insights articles – “The Titans are Coming” & "The Post-Attention Era is Coming", these will change the fundamental architecture of how data is stored and retrieved by AI models in 2026 with the net result of better AI.

In particular our focus in 2026 will be on data sovereignty, governance and ethics. We are already advising our clients to take control and ownership of data to prepare for the governance that is coming in 2026. On the 2^nd of August all articles (except for Article 6: Classification Rules for High-Risk AI Systems) of the EU AI Act comes into force and along with other ISO standards and global governance, the need for explainable AI will rely on robust data management.

As we talk about governance, we should address the public outages of critical tech services in 2025. Cloudflare being the most notable highlighting how crucial the Edge network is in its growing importance in parallel to cloud, but over the year each cloud service providers had their issues critical services. The bad news is that is going to be more common as the ‘Time to Market’ metric continues to trump reliability or resilience metrics. The more worrying sign is that end-users are growing apathetic to this rather than demanding better, resulting in the normalisation of global outages as ‘just one of those things’. This coupled with the earlier mentioned decline in code quality, lingers ominously for the industry in 2026. This will need to be addressed sooner or later and in 2026 we will be doubling down on sensitive data on secure private infrastructure, dynamic scaling across cloud services and inference at the edge. The architecture patterns for this across hybrid-cloud and edge-compute infrastructure are well established and in 2026 we expect AI to embrace these best practices to mature into different categorisations of AI models that are specific to the parameters of their use-cases.

That brings us to our next prediction for 2026. Smaller is better. The large general purpose models are overkill when the majority of products just need a LLM to understand basic context not the sum of human knowledge, a little specialised knowledge in a domain is better than a lot of general knowledge in domains not relevant to the use-case. When you factor in the costs this becomes even more obvious so at JP Global we use multiple smaller models instead of a large model and we have found this to be orders of magnitude more efficient and effective. We think more and more people are going to find this and invest in using smaller models. This is fuelled by the rise of more Open Source models. We have yet to feel the effects of the Chinese State Department strategy of Open Sourcing AI  but 2026 is the year that the closed-source models will start to feel the tremors of existentialism from Open Source alternatives. China is also leading the charge in the research and development area of AI. NeurIPS had an unprecedented year with double the number of AI related paper submissions to previous years and in 2025 24% of the 21,575 were accepted, another record. The trends point to a maturing of the technologies under the hood of AI and this bodes well for the future of AI as the geopolitical implications of AI are felt at a societal level.

Looking To The Horizon

As we look further ahead than 2026 to stay ahead of emerging tech trends. AI and the paradigm shift that brings pales in comparison to the paradigm shift that quantum systems will bring. In 2026 we are expecting the trend of small but crucial tangible progress made in system coherence, error correction, novel material science, information processing etc. Quantum is already queued up as the next big hype buzzword and undoubtedly you will hear more and more about it in 2026. Our focus remains on the fundamental science and engineering of quantum systems for now before we scale quantum systems to solve unique industry problems.The final word is that we at JP Global are bullish on the industry and will be sharing a lot more about each of the topics touched on here over the course of 2026. Meanwhile, if you want to know more get in touch, we would love to hear from you and you can claim a free coffee from us for getting to the end of the article.

Happy 2026!