Midyear 2026 Check-In: AI, Supply Chains, and the NDC-12 Story Nobody's Watching

When I spoke with PharmTech in late 2025 about what I expected for pharma manufacturing in 2026, I spoke to the drivers of innovation and how drug manufacturers should prepare their operations for the next phase of competition and growth. Now that 2026 is half over, I reflect back on what predictions I made came to fruition, which didn't, and why, as well as what's in store for the second half of the year.

AI in R&D and the Shift to High-Value Therapies

The AI-driven R&D trend I described in December continued to accelerate in the first half of 2026, but with an important nuance I didn't fully anticipate. In December, I characterized AI's role in drug discovery as primarily about cutting timelines for identifying candidates. What's become clearer in the first half of 2026 is that the more meaningful shift is happening at the strategy level, not just the speed level. Companies are discovering candidates faster and restructuring their pipelines around what AI makes economically viable to pursue. The oncology and rare disease focus has intensified, partly because those are the areas where AI's ability to find signal in complex genetic data has the most commercial leverage.

Generative AI for de novo molecular design, which I flagged as a 2026 driver, is progressing but more slowly than the January headlines suggested. The technology works in controlled research settings. Getting it through regulatory scrutiny is a different challenge, and the interpretability problem I mentioned in December as a “black box” concern is still the central obstacle for anything approaching a regulatory submission. The talent shortage I highlighted has, if anything, gotten worse. People who understand both the biology and the AI are genuinely rare, and everyone is trying to hire the same small pool of people. Real-world evidence in regulatory submissions is moving, but modestly. FDA has been more receptive than expected in some therapeutic areas, less so in others. It's not the clean breakthrough moment I anticipated and this is a negotiation happening on a case-by-case basis.

Supply Chains and Global Disruptions

This is where the industry's trajectory aligned most closely with my earlier outlook. I said in December that the sector was moving away from purely cost-efficient, just-in-time models toward genuine resilience through diversification. What I didn't fully anticipate was how decisively the tariff environment in early 2026 would force that transition from a strategic option to an operational necessity. The “China+1” framing I used was already too conservative by March. I'm seeing companies building genuine multi-region redundancy, not just adding one backup supplier and calling it done.

US manufacturing investment increases, which I flagged as a geopolitical hedge, have become a strategic priority rather than a contingency plan for generic manufacturers. The investment decisions being made now are real and significant. The companies that started this shift two or three years ago are in a materially stronger position today.

One thing I underestimated: the supply chain resilience conversation is no longer primarily about geography. It's increasingly about data visibility. You can diversify your supplier base, but if you can't see what's moving through it in real time, you still can't respond to disruptions quickly enough. The companies that combined geographic diversification with genuine serialization and traceability data infrastructure are responding to disruption faster than those who did only one of the two.

Competitive Advantage as Margin Pressures Intensify

The high-barrier-to-entry product focus I described in December is playing out, but the competitive dynamics have shifted in ways worth noting.

Complex injectables and biosimilars remain the right strategic focus, but the competitive pressure in those categories has intensified faster than most anticipated. Several companies that were counting on complexity as a moat are discovering that complexity alone isn't sufficient and that execution excellence matters as much as product selection. The manufacturers who are winning combined the right product portfolio with supply chain infrastructure that can deliver those products reliably at scale. Complex products with unreliable supply chains don't generate the returns they should.

Strategic M&A remains active, as I predicted, but the integration challenge is where I've seen the most difficulty. Companies are acquiring the right pipelines but not underestimating what it takes to absorb a new product family into an existing serialization and regulatory compliance architecture. That's not a technology problem; it's a governance and process problem that technology alone doesn't solve.

AI in Manufacturing and Smart Factories

The prediction I was most optimistic about in December—agentic AI would begin entering production manufacturing environments—is having a mixed response, if not running behind. Cash flow-rich and branded organization adoption is at a high rate. In a GxP environment, you cannot deploy something and iterate. Every change to a system that touches product release must be validated, documented, and auditable years later. That's not a technology problem; it is the right constraint for an environment where errors reach patients. But it means the “year of the agent” framing that got a lot of coverage at the start of the year has progressed more slowly than many early-2026 predictions suggested, particularly in regulated pharmaceutical manufacturing.

What did happen and what I think was underreported is that data quality became the central competitive differentiator for AI in manufacturing, even faster than I expected. I said in December that AI barriers included data fragmentation and interpretability. What I didn't say clearly enough is that data quality isn't just a barrier; it's the new moat. The companies that are realizing real AI value in early 2026 are almost uniformly the ones that invested in their ERP, serialization, and master data infrastructure in 2023 and 2024. The ones still in pilot mode tried to layer AI on top of fragmented systems and discovered that the problem was never the algorithm. When AI begins processing thousands of supply chain events and making recommendations, inaccurate data causes errors, and at machine speed. It's a “garbage in, garbage out” scenario, but now it happens before anyone can catch it manually.

On smart factory upgrades specifically, I said in December that manufacturers should plan significant investment in IoT sensors, robotics, advanced automation, and cloud infrastructure to reach Industry 4.0 standards. That investment is happening, but the organizations getting the most from it are those that treated it as an operating model transformation rather than a technology installation. You cannot bolt AI onto a fragmented factory and expect it to work. The digital factory cannot outperform poor enterprise data.

The NDC-12 Story I Didn't Fully Anticipate in December

I think this will be the supply chain and regulatory story of H2 2026 and into 2027, and most coverage is still treating it as a packaging update. It isn't.

NDC-12 is an enterprise identity transformation. When the FDA's 12-digit National Drug Code rule fully phases in, virtually every small or large enterprise system in the pharmaceutical supply chain that stores, processes, transacts, or reports a drug product identifier will be affected: ERP, serialization, EPCIS data exchange, EDI transaction formats, warehouse management, labeling, chargeback platforms, reimbursement systems, contract manufacturer connections, distributor onboarding, etc. All of it.

Through my participation in the cross-industry NDC-12 Pilot Program and EDI integration workshops, I've heard conversations that make clear that the scope of what needs to change is significantly larger than most organizations currently appreciate. The companies beginning their systems assessment now will be in a very different position from those that wait until 2031 and try to do it reactively.

The parallel I keep using is DSCSA. The companies that treated DSCSA as a genuine operational transformation—building clean serialization architecture, establishing real governance, getting trading partner data quality right—are now in a much stronger position for NDC-12 and for every AI initiative they're running. Those that treated DSCSA as check-the-box compliance are rebuilding from weak foundations. History is about to repeat itself for organizations that make the same mistake with NDC-12.

Perhaps the biggest lesson from the first half of 2026 is that organizations are discovering technology is no longer the limiting factor. Most enterprise platforms already support AI, automation, IoT, serialization, cloud computing, and advanced analytics. The competitive differentiator is increasingly organizational readiness with trusted and high-quality data, governance, cross-functional collaboration, workforce capability, and disciplined execution. Technology has become accessible; transformation has not.

What to Expect for H2 2026

Three things stand out.

1. DSCSA Enforcement Reality Arrives

As the DSCSA stabilization period concludes, organizations should expect greater operational scrutiny from trading partners and increased emphasis on serialization data quality, interoperability, and exception management. Serialization data accuracy and trading partner interoperability are enforceable requirements now, not performance aspirations. The first signs of this are already visible in how distributors are responding to manufacturer compliance metrics. Another sign is that regulators have been emphasizing DSCSA compliance and performance while they visit the enterprises for inspections and audits. Right documentation, quick explosibility, and provability is still a hard part for many in the industry.

2. NDC-12 Systems Work Starts in Earnest

The deadline is 2033, which feels distant. The systems changes required do not have 2033-length lead times. EDI format migrations, master data architecture changes, EPCIS updates, trading partner re-onboarding, and more all have lead times measured in years. The next few months of 2026 is when I expect the gap between organizations that are in their assessment phase and those that will start to become visible in commercial relationships, not just compliance conversations.

3. AI Finds its Most Defensible Pharmaceutical Use Case

My prediction is that AI applied to serialization data and supply chain exception handling will show the clearest, and one of the earliest, enterprise-scale AI use cases capable of delivering measurable ROI, because the underlying data foundation already exists. Serialization systems generate granular, structured, event-level data on every product unit through the entire supply chain. Applying AI to that data for anomaly detection, exception prioritization, and compliance risk identification doesn't require building a new data foundation. Instead, it requires applying the analytics layer to one that already exists and is already validated. That's a materially shorter path to measurable outcomes than most other AI applications in pharmaceutical operations.