Tag Archive for: artificial intelligence

2026 Predictions for Semiconductors: AI, Chiplets, and the Path to Sustainable Innovation

As we step into 2026, the semiconductor industry stands at the crossroads of unprecedented technological advancements and complex global challenges. From the rise of AI-driven chip design and heterogeneous integration to the growing emphasis on sustainability and geopolitical shifts, the sector is navigating a transformative era.

The next wave of innovation will be defined by breakthroughs in advanced lithography, chiplet architectures, and quantum computing, while sustainability efforts will reshape manufacturing processes to address energy efficiency, water usage, and materials recycling. At the same time, the industry faces critical hurdles, including talent shortages, supply chain realignments, and the need for robust cybersecurity measures.

In this year’s predictions series, we’ve gathered insights from leading semiconductor experts:

• Simon Bennett – Principal Solutions Consultant, AI Tech Sales
• Sarah Crary Gregory – Principal Engineer, Systems and Solutions Methodologies, Crary Labs
• Neil Stroud – General Manager Automotive and Semiconductor, Jama Software
• Steve Rush – Principal Consultant, Jama Software

Together, they explore the trends and technologies shaping the future of semiconductors. From AI-driven automation and edge computing to the challenges of regulatory shifts and the promise of chiplet-based architectures, this piece highlights the innovations and strategies that will define 2026 and beyond.

Curious about what’s happening in other fields? Read part one on consumer electronics, part two on medical device & life sciences, part three on aerospace & defense, part four on automotive, and stay tuned for our upcoming predictions for AECO.

1: Emerging Technologies

Q: What emerging technologies (e.g., advanced lithography, AI-driven chip design, quantum computing, heterogeneous integration) will have the most transformative impact on the semiconductor industry in the next five years?

Simon Bennett: In the next five years, the semiconductor industry will continue to grow, almost doubling in size from today to $1Trillion by 2030. But to sustain that growth, the industry will go through some extreme changes and challenges. The first trend to note is actually due to a declining trend as Moore’s Law continues to slow. [Editor’s note: Moore’s law is the observation that the number of transistors in an integrated circuit (IC) doubles about every two years.]

Moore’s Law has driven the growth of the Semiconductor industry for many decades, but it is bumping up against the fundamental laws of physics. The economics of scaling to the next node are increasingly prohibitive and taking longer and longer to reach fruition.

Whilst keeping an eye on what is coming out of China, there will be some more mundane but equally challenging technology trends that are emerging and will become increasingly important in 2026 and beyond. These are AI driven design, and both chiplet and wafer scale designs (two opposite ends of the spectrum, but both an engineering reaction to the slowing of Moore’s Law).

Neil Stroud: Given the ever-increasing innovation around AI and its associated deployment, chip development is under continued pressure to keep up. This is applicable across all architectures, including Central Processing Units (CPUs), Graphics Processing Units (GPUs), and Neural Processing Units (NPUs). Naturally, continued optimization will happen around acceleration and emerging technologies like process node shrinks (advanced lithography), AI-driven chip design, and the chiplet approach (heterogeneous integration). Process node shrinks will contribute. However, the chiplet approach will also drive heterogeneity across architectures and nodes. All these factors will intimately impact the next generation of chip families for AI in the datacenter and at the edge.

2: Sustainability and Manufacturing Efficiency

Q: How do you see sustainability influencing semiconductor manufacturing, particularly in areas like energy efficiency, water usage, and materials recycling? What strategies will help the industry achieve greener fabrication processes?

Bennett: This is a great question, and right now, the elephant in the room. From Fabs to datacenters, the environmental impact is huge. Water consumption alone is a huge factor. Twenty years ago, visionary realtors quietly purchased acres of land close to a bountiful supply of water and close to a large data pipe. Those realtors are now wealthy, and the secret is out. Now the price of that land is at a premium. So, the investors behind the fabs and the datacenters are using government subsidies and their own funds to find alternative sources of energy and resources. Nuclear is making a comeback, driven in part by the energy demands of the datacenters. Municipal areas like Phoenix are making guarantees of plentiful water to companies to attract them to their region; that will put them in direct conflict with farmers in California.

Most of this is happening off the radar of the mainstream media, and the political arena is presented as a battle for the best jobs. The concern over the environmental impact is not yet front and center. Two events will likely happen to change this:

The AI bubble will inevitably burst. Just like in the early days of the internet, there will be market correction as reality catches up to expectation. Just like the internet bubble, this doesn’t mean that AI is not going to be a societal change; it just means the market got too overheated.

Unfortunately, there will be some kind of accident related to the overbuild of the infrastructure around Datacenters and Fabs. A dam will burst (Phoenix – see Roosevelt Dam), or a multibillion fab will be damaged by a natural disaster (see fault lines in Taiwan). These two events will raise awareness of environmental costs relating to sustainability and manufacturing efficiency.

In other words, in the next five years, we will be forced to take a pause, a breath, and truly measure the value vs the cost. This isn’t a bad thing. Our human history of technology transformations is punctuated with these pauses and resets. Usually for the better.

Steve Rush: Sustainability is hugely influential and important. Energy demand is forecasted to accelerate with new data centers and the demand for AI. Semiconductor companies need a system to help manage their sustainability requirements and, very importantly, validate them. Implementation to hit targets and balance, power, efficiency, and sustainability will be a series of trade-offs – semiconductor organizations will need a tool to trace all of this information and prove that they meet sustainability targets and goals.

Sarah Crary Gregory: While the semiconductor industry is obviously fiercely competitive, it can match that intensity with fierce collaboration on critical issues. Sustainability is probably the most prominent area where industry consortia such as the Semiconductor Climate Consortium bring companies together to tackle common problems. Initiatives to enable water reclamation, reduce emissions, and produce data quantifying the return on investment of sustainability practices will be more critical with the burden placed on these resources from the exponential expansion of AI. The semiconductor industry is highly interdependent, and nobody believes that there’s a way to get a competitive advantage by monopolizing natural resources. The way forward is through innovations that decrease resource consumption and minimize waste, and initiatives for water reclamation/”net zero” resource use will continue to be essential investments.

Stroud: I think there are two parts to this. Firstly, the environmental impact of actually building the chips in foundries. A huge amount of effort and investment has gone into sustainability in semiconductor manufacturing, including energy efficiency, water usage, and materials recycling. semiconductor manufacturing and materials. A great example of this is massive recycling of water used in fab processes, as well as optimizing processes and the associated chemicals used, including minimizing atmospheric emissions.

Secondly, there is the environmental impact related to the deployment of the device itself, as it consumes power and emits heat. Of course, the extreme example of this is the data center where huge racks of GPUs or CPUs are deployed, collectively consuming Megawatts of power to both power them and cool them. Again, huge investment is going into driving data center efficiency. One way to contribute is through chip design optimization to improve ‘performance per Watt.’ That is simply a measure of how much computing can be done for a given Watt of power. This optimization can happen through design and architecture efficiencies as well as process node shrinks. Ensuring the software stack is also developed to drive efficient use of the underlying hardware platform also has a fundamental role to play. It’s easy to see that these steps can have a profound positive impact on the environment caused by the global electronics footprint.

RELATED: Buyer’s Guide: How to Select the Right Requirements Management and Traceability Solution

3: AI and Automation

Q: How is AI accelerating innovation in semiconductor design, verification, testing, and manufacturing? What challenges must companies overcome to fully leverage AI-driven automation?

Bennett: Natural language and agentic AI will continue to show up across the tool chain. But expect some resistance from SOC design engineers, who, ironically, since they are at the epicenter of the AI revolution, are traditionally conservative and slow to adopt new methods. Verification is the most in need of help with AI-driven automation, since there just aren’t enough engineers on the planet to drive the verification needs of an SOC. (see salaries on Glassdoor). It’s been estimated that with the use of AI, a team of 3 expert verification engineers can do the work of 5 traditional verification engineers with limited use of AI, in 3 to 5x less time. This is a compelling message to an (open-minded – see below for a caveat) engineering VP struggling to find the resources to deliver a fully validated product on time. These engineers and the tools they use will be in high demand in the next five years.

Beyond design, AI will show up in yield and manufacturing analytics. The challenge of inventory and yield management in the era of disaggregated chiplet-based designs is magnified. It’s essential that all the chiplets deliver the yield and volume needed at the exact same time. The overall package is only as good as the weakest tile. This is an underserved opportunity within the big three EDA companies, and the packaging OEMS tend to jealously protect their homegrown investments in solving these challenges. Expect emerging startups to come forward as disruptors in this particular segment in 2026 and beyond.

Rush: Every company is looking for ways to utilize AI in their organization. AI can play an important role in managing traceability, especially from siloed systems that are isolated from one another. Agentic experiences that improve engineer productivity really are key. The main challenge that AI has in the semiconductor space, in particular, is adoption with the engineering team. AI experiences must improve engineering productivity; they must be accurate, and they cannot be an impediment to use. If AI-generated content is of questionable quality or if the AI experiences become too burdensome to use, AI initiatives risk dying on the vine.

4: Supply Chain and Geopolitical Shifts

Q: How are global supply chain realignments and geopolitical factors shaping semiconductor strategy? What can companies do to mitigate risk and ensure resilience in developing complex products on their own or with co-development partners?

Bennett: A global supply chain developed over the past thirty years has delivered $1T in cost savings. This $1T is now under serious threat as the world is a very different place compared to when this globally interconnected environment was first conceived. In the next five years, expect China to become more self-sufficient as it replicates every aspect of what it previously relied on from overseas, from EDA to IP to fab equipment. Expect to see semiconductor-based products from coffee machines to phones to servers to (even) EVs sourced almost exclusively from China with little to no reliance on anything beyond the shores of China. This will trigger protectionist measures in the US and the EU as they work to protect homegrown industries from what will become increasingly consumer appealing products from the Chinese factories.

A more optimistic view may be that the tensions ease as the US / EU recognize the need for open trade with China, and continue to see its designs realized in Chinese factories (but I’m not holding my breath). In semiconductors, companies will be most susceptible to this shift in China as they move to homegrown alternatives. As the geopolitics ramp up, the focus on Provenance in the West will become a C-suite / US Senate / EU Parliament level of attention. Knowing where every component or piece of code originates, its genealogy will become paramount. A counterforce will emerge where the information is “buried” as the realization hits that we can’t possibly trace the root of every bit of code, every nanometer of design. Companies will emerge with one of two unique value propositions: 1) we can audit your product and provide the provenance, 2) everything you use is contaminated; we are a new company, built cleanly from the ground up. Somehow, all three will survive – the traditional companies, the auditors, and the new “clean” companies. But there will be some very interesting mergers and acquisitions, mostly off the radar as these three entities re-align and learn to co-exist.

Rush: These days, you can basically count on major geopolitical news covering the semiconductor industry week in, week out. At the end of the day, co-development and partnerships are key. The semiconductor supply chain is mind-bogglingly complex. Adopting modern, more collaborative tooling is on the rise. Historically, the semiconductor industry has even been hesitant to adopt cloud-based solutions, and I’ve definitely seen a change in the last few years around this.

Stroud: Like many other segments, the semiconductor market tends to be cyclic, which leads to times of undersupply and oversupply. This is a complex problem to manage with many factors, including global supply chain realignments and geopolitical factors. Naturally, foundry capacity has a big role to play, and we seem to be in an investment phase right now with a number of fabs being built. This is a massive investment with a modern fab costing tens of billions of Dollars and taking multiple years from construction start to mass production. Communication and collaboration across the ecosystem also has a role to play, especially now that we are accelerating into the chiplet era, which can help mitigate risk and ensure resilience in developing complex products.

5: Chiplet and Heterogeneous Integration

Q: What role will chiplet architectures and heterogeneous integration play in addressing performance and scalability challenges? What technical and ecosystem hurdles must be overcome?

Bennett: Chiplets are essential to the continued growth of Semiconductors. Without chiplets, the forecast CAGR ($1T by 2030) is unreachable (basic economics of Moore’s Law). The challenges are two-fold: 1) engineering challenges around interconnecting tiles from different suppliers running at high speed and with the thermal challenges of a modern chip; and 2) coherence – the coherence of the supply chain, compliance, and verification. More specifically, the standards emerging need to be better governed (e.g., Universal Chiplet Interconnect Express (UCIe) for interconnect and system architectures if they aren’t going to become bottlenecks stymying growth.

6: Talent and Workforce Development

Q: With growing global demand for skilled engineers and manufacturing specialists, how can companies address the talent shortage in the semiconductor industry?

Bennett: This is where AI needs to step in and become more readily accepted within Semiconductor Engineering orgs. As stated above, studies show that a small team of AI proficient verification engineers are 5x+ more productive than a traditional team. However, the resistance comes from within – engineers are conservative, and within a traditional engineering organization, the manager / Director / VP still measure their worth by the number of engineers the corporation is willing to fund. This leads to destructive behaviors, such as a VP of Verification Engineering employing 100 RTL validation engineers to do the job that 10 Functional Verification engineers could do because “it’s too expensive to hire the functional verification engineers” – the companies that will thrive and succeed in the next five years are the ones who break down this cultural impasse.

Rush: There are a lot of talented people in the job market right now who can help fill the gap. Hopefully, semiconductor companies will look to hire talent from across industries – automotive, medical, and aerospace. There are certain challenges in getting enough skilled foreign workers to fill certain roles – I’m more concerned that there are many highly skilled, talented people out there looking for jobs!

7: Regulatory and Export Controls

Q: How do evolving export controls, trade policies, and security regulations impact semiconductor innovation and competitiveness? How can companies adapt strategically?

Bennett: They don’t impact semiconductor engineering innovation or competitiveness – in fact, they improve it. Case in point is China – as access to advanced GPUS / EDA tools was limited, they innovated, and actually improved on the technologies they didn’t have access to. Another example is where the Russian engineers working for US companies prior to the war in Ukraine were let go and went to work for Russian companies, helping boost the AI business in Russia. But where the question applies is the innovation at the corporate level. Engineering innovation can be stymied by a C-suite overly concerned about trade or political issues. The paradox is that smaller companies with less of a global or political reach could feel less compelled to avoid the risk associated with innovation.

Gregory: “Evolving” is an understatement! The volatility around export controls and trade policy in the United States right now is simply unprecedented, and 2026 looks like more of the same. Companies can strategically navigate these unsettled times by implementing systems –people, processes, and tools – that enable maximum response flexibility. Modular architectures, whether they’re chiplet-based, specific configurations of IP cores, highly modular software, or other building blocks, will enable the development and delivery of products whose configurations can be changed and modified as circumstances warrant. Variant management is a critical capability to be able to swap features in and out based on policy changes. Solid, well-governed data foundations will be critical to stay on top of the wildly shifting policy landscape.

RELATED: Engineering Governance is a Critical Business Strategy for Product, Project, and System Development Excellence

8: AI and Edge Computing Demand

Q: As demand for edge AI and high-performance computing grows, what innovations are most critical to meet performance and power efficiency goals?

Bennett: There are many ways to answer this, but I’ll focus on the chip-level design aspect. First, the interconnect, as previously described – the clean adoption of UCIe and a strong governing body to oversee its evolution (think Universal Serial Bus, or USB.) 3D packaging needs to keep up with the thermal demands of a heterogenous package – this may lead back to the engineering talent pipeline previously discussed since the engineers who have the combination of skills to analyze and design (future-proof) these packages are unique (think warping of a substrate as it reacts to thermal pressures, leading to subtle issues with the interconnect manifesting as signal integrity.)

Rush: I’ll answer this more from a – data isolation – perspective. Design and testing are really important, but more important is tracing all the way to the highest level and validation. I think responsible AI will help with efficiency here, but companies need a way to trace from the top down. In all honesty, this is a challenge for the semiconductor industry – having one single source of truth that can prove you’re hitting sustainability goals.

9: Cybersecurity and IP Protection

Q: With increasingly complex global supply chains, how can semiconductor companies protect intellectual property and secure their design-to-production ecosystems?

Bennett: Expect a lot more reference to initiatives such as Software Bill of Materials (SBOM) and Engineering Bill of Materials (EBOM.) Expect the concept of a Bill of Materials (BOM) to evolve and take on more significance in the next few years. Expect the term Provenance to take on more importance. Traditional PLM companies will position themselves as the answer, but there will be significant pushback from the semiconductor industry, and rightly so – these PLM systems were never developed with semiconductors in mind. They are monolithic in nature, expecting the end user to move their data into their environments. The C-Suite will sign on, the engineers won’t. This will lead to QMS and IT organizations emerging to manually clone the data inside the PLM systems. For a while, this will seem just fine, until one or more issues come to public light, and the C-suite exec realizes they have spent a lot of money on tools and resources, and it didn’t solve the problem. Those companies that invested in a more lightweight engineer-friendly solution, providing traceability, compliance, and coherence insights without the costly overhead of monolithic tools and the resources that go along with them, will grab the attention of those who lost out. And yes, AI will play a part. A well-managed digital thread with the ability to expose itself in a controlled manner to intelligent insights will win out.

Rush: I mentioned earlier that semiconductor companies are adopting more cloud-based tooling. But they are not slacking in terms of security needs. By selecting best-in-class tools with exceptional infosec track records (like Jama Connect), they are effectively balancing speed and agility with security and not sacrificing either. They are pushing their vendors to expand their tool sets to deliver best-in-class experiences with rationale, scalable permission structures that are tightly governed. They’re looking for tools and vendors that are putting AI at the center of their vision – but need their vendors to offer closed, secure LLMs or integrations with in-hours AI systems.

Stroud: This is not a new issue! The semiconductor industry has been wrestling with intellectual property protection and securing the design-to-production ecosystem for years. The challenge is how to build enough flexibility in the ‘fixed’ silicon that, when combined with software (across all layers), is able to guard against future exploits and vulnerabilities. It’s almost impossible to build a modern chip without multiple integrated security capabilities. Also, it’s worth noting that security has to be a multidimensional approach in this age of hyperconnectivity, spanning seamlessly from cloud to edge. This is why we see an ever increasing number of emerging security standards that apply to both implementation and development processes, impacting hardware, software, and system design and deployment.

10: Future Outlook

Q: What do you see as the most important technological and market shifts that will define the semiconductor industry five to ten years from now? How can companies position for sustained leadership?

Bennett: 1) Semiconductor Technology: Chiplets, and the packages that are needed to realize their promise to alleviate the decline of Moore’s Law. 2) Companies: very different answer–the companies that will succeed in the future are those that completely obfuscate the hardware considerations from their customers—it’s all software, don’t worry about the hardware – we have taken care of that.

In summary, in some ways it’s the same old story – recognize and reward the unique engineering talent that helps differentiate your product, understand what the customer wants, and remove the barriers to growth. Sounds simple, right?

Rush: With AI, the amount of data that companies will manage is going to increase tremendously. Trying to manage that traceability is going to be extremely challenging. Jama Connect, with the new scaling improvements and AI vision, is at the forefront of the market and uniquely positioned to help semiconductor companies here.

Gregory: Agreed. AI is already reshaping the demand side of the market equation. The supply-side will evolve to support highly customized semiconductor design, even purpose-built and assembled solutions that are rapidly defined and fabricated. Edge AI and NPUs (neural processing units), along with open architectures such as RISC-V (and the RISC SW Ecosystem), will further broaden the horizons for semiconductor companies. How to be positioned for success? Again, it’s all about response flexibility. Sensing both strong and weak signals in the market and systematically building resilience into the company’s organizational practices will determine which companies emerge stronger from the challenges of the next five to ten years.

Intelligently improve your development process with Jama Connect:
Start your free 30-day trial!