UP–NS Rail Merger: How AI Helps Shippers Prepare

Union Pacific and Norfolk Southern are putting a date on the calendar: Dec. 19, 2025 is when they say they’ll file a formal merger application with the Surface Transportation Board (STB) for an $85 billion deal. If approved, it would create the first coast-to-coast, single-line transcontinental railroad in the U.S., spanning 53,000 miles of track across 43 states.

Most coverage will focus on the politics and the regulatory fight. That’s fair—this will be one of the biggest transportation decisions of the decade.

But if you’re a shipper, 3PL, broker, or manufacturer, the practical question is simpler: How do you plan your network when the network itself may change? This is exactly where AI in transportation and logistics earns its keep—because the winners won’t be the people with the strongest opinions. They’ll be the teams with the best scenarios, cleanest data, and fastest ability to re-route when reality hits.

What’s actually being proposed—and why it matters operationally

Answer first: The UP–NS deal is pitched as an “end-to-end” combination that replaces interchanges with single-line routing, potentially reducing transit time and paperwork on certain lanes.

UP and NS say a combined network can shave up to two days off freight moving through major interchange chokepoints like Chicago and St. Louis. The logic is straightforward: fewer handoffs means fewer chances for dwell time, switching delays, missed connections, and administrative friction.

The big promise: fewer handoffs, fewer surprises

Rail interchanges are where plans go to die.

Even when service is “normal,” interchange performance is variable. When service is strained—peak season, weather events, labor constraints—interchanges amplify delays. A transcontinental, single-line route could:

• Reduce interchange dwell on some lanes
• Simplify exception management (fewer parties to coordinate)
• Cut admin workload (contracts, claims, status checks, reconciling multiple systems)

The big fear: consolidation and merger disruption

Answer first: Shippers opposing the deal are worried about reduced rail options and the risk of service instability during integration.

Those concerns aren’t abstract. Past major rail mergers created periods where operations didn’t just get slightly worse—they got unpredictable. And in logistics, unpredictable is often more expensive than slow.

What changes after a merger isn’t just branding. It’s dispatching rules, crew starts, terminal operating plans, car management practices, network priorities, and—critically—IT systems.

The STB timeline: what the next 12–18 months could look like

Answer first: The filing starts a multi-stage STB process: a 30-day completeness review, then public acceptance or rejection, followed by 45 days of comments, 90 days for responsive applications, and then an official review likely lasting a year or longer.

For logistics teams, the regulatory timeline matters because it defines your planning rhythm.

Here’s how I’d translate the process into business planning checkpoints:

1. Now through the completeness check (next ~30 days): Treat this as the “facts gathering” window. You’re building your baseline model.
2. Comment and response periods (next 2–4 months): Expect new details, counterclaims, and shipper coalition positions. This is where lane-level risk assumptions get refined.
3. Official review (12+ months): This is where you run scenarios like you mean it—because uncertainty will linger, and commercial behavior can change even before formal approval.

A merger review isn’t just paperwork. It’s a long period where carriers, customers, and competitors start adjusting their moves.

How a transcontinental railroad could reshape shipper network design

Answer first: If UP–NS is approved, the biggest network impact will be on east–west intermodal and carload lanes where today’s service depends on interchange performance and terminal congestion.

This is where supply chain leaders should be opinionated: don’t wait for a decision to start planning. Build options now.

1) Intermodal routing gets simpler—until it doesn’t

Single-line intermodal moves can reduce coordination overhead, but they also increase dependence on one provider’s network health. If a merged network prioritizes certain corridors, some lanes could improve while others degrade.

Practical implication: your “best” lane today may not be your best lane six months after integration.

2) Chicago and St. Louis: less interchange isn’t the same as less congestion

If fewer handoffs occur in Chicago and St. Louis, some freight may bypass classic interchange pain. But terminals still have physical constraints—tracks, lifts, crews, slots.

What changes is the shape of congestion:

• Some interchange delays may shrink
• Some terminal queues may move elsewhere
• New bottlenecks can appear where networks connect to customers, ports, and ramps

3) Procurement strategy will shift from “rate shopping” to “resilience shopping”

Consolidation tends to tighten choices on certain origin-destination pairs. When your option set shrinks, your negotiating leverage shifts—and the premium you’ll pay is often for service commitments and recovery options, not just price.

If you’re planning 2026 and 2027 bids, it’s time to separate:

• Lanes where rail is a tactical spot buy
• Lanes where rail is a strategic asset with contractual performance terms

Where AI in transportation and logistics fits (and where it doesn’t)

Answer first: AI is most valuable here for scenario planning, predictive ETA risk, and network optimization under constraints—not for guessing whether regulators approve the deal.

I don’t trust “AI predictions” about regulatory outcomes. The useful work is different: using AI to reduce the cost of uncertainty.

AI use case #1: Scenario planning you can actually execute

Build 3–5 merger scenarios, not 50.

Good scenarios are specific enough to price and route:

• Scenario A (status quo): no merger, competitive landscape holds
• Scenario B (approved + smooth integration): faster single-line routing on defined corridors
• Scenario C (approved + integration disruption): higher variance in dwell and missed connections for 6–12 months
• Scenario D (approved + remedy conditions): mandated open gateways, divestitures, or service commitments

AI helps by turning those scenarios into lane-by-lane playbooks:

• Which SKUs can tolerate extra variability?
• Which customers require tighter OTIF?
• Which facilities need added buffer stock if rail variance rises?

AI use case #2: Predictive risk scoring for rail ETAs

Rail planning breaks when you treat every shipment like it has the same risk.

A practical AI approach is a shipment-level risk score based on:

• Lane history (dwell, missed connections, variability)
• Terminal performance signals
• Equipment availability patterns
• Calendar effects (holidays, peak shipping weeks)

Even without “perfect” real-time rail visibility, you can model the probability of delay and take action early:

• Rebook to alternate ramps
• Switch mode (truckload / dray + transload)
• Pull orders forward
• Adjust appointment scheduling with customers

AI use case #3: Mode and route optimization under consolidation

When options shrink, optimization gets harder because constraints matter more:

• Limited interchange gateways
• Equipment type requirements
• Service windows and delivery appointments
• Demurrage and detention exposure
• Customer penalty structures

AI optimization earns value when it produces decision-quality recommendations, not pretty dashboards.

A useful output looks like:

• “For these 14 lanes, shift 20% volume to alternative rail or truck during weeks 3–8 of integration risk.”
• “Hold safety stock at these 3 DCs because predicted rail variability exceeds your service tolerance.”
• “Renegotiate contracts on lanes where you’ll likely lose competitive routing optionality.”

What to do now: a practical prep checklist for shippers and 3PLs

Answer first: Treat the merger as a live risk-and-opportunity program: map exposure, build alternatives, and put data discipline in place so you can move fast.

Here’s what works in the real world.

1) Map your exposure (in one afternoon)

Pull the last 12–18 months of shipments and label:

• Lanes touching UP or NS networks
• Moves relying on interchange in Chicago or St. Louis
• Top customers by penalty/chargeback exposure
• High-value or time-sensitive SKUs

If you can’t do this quickly, that’s the first problem to fix.

2) Build a “routing options” library before you need it

For critical lanes, document at least two alternatives:

• Alternate rail routing (different ramps/gateways)
• Transload strategy (rail to truck or truck to rail)
• Pure truck fallback (including capacity assumptions)

This is boring work. It also saves careers.

3) Set performance baselines now (so you can detect change later)

If the merger progresses, everyone will claim service improved or worsened. Your team needs its own truth.

Baseline metrics:

• Door-to-door transit time (median and 90th percentile)
• Dwell time by terminal (where available)
• On-time pickup and on-time delivery
• Claims frequency and cost
• Accessorial trends

4) Align procurement, operations, and customer service

Merger uncertainty punishes silos.

If procurement pushes volume to hit a rate target while operations is fighting service fires, you lose twice: cost and reliability.

Create a joint operating stance:

• Which lanes prioritize cost?
• Which lanes prioritize reliability?
• What’s the trigger to switch modes?

5) Decide where AI automation is safe—and where humans must stay in the loop

Automate repeatable decisions (rebooking rules, exception triage), but keep humans on:

• Contract changes and service commitments
• Customer communications during disruptions
• Volume reallocation across modes when capacity tightens

A simple rule: if the decision creates customer promises, don’t fully automate it.

What I expect next—and what I’d watch closely

Answer first: Expect the public narrative to focus on “competition” while the operational reality hinges on terminals, intermodal ramps, and how integration is sequenced.

If you’re tracking the deal’s logistics impact, watch for:

• Commitments on open gateways and interchange access
• Service assurance plans (staffing, terminal ops, recovery playbooks)
• IT integration approach (the quiet cause of loud failures)
• Any early commercial behavior shifts—rate changes, capacity allocation, lane prioritization

And remember: even if the merger isn’t approved, the process itself can change behavior. Competitors respond. Shippers rebid. Networks adjust.

Rail is long-cycle infrastructure. You don’t wait until the concrete dries.

Most logistics teams already use AI for forecasting and transportation planning in some form. The next step is using AI for network uncertainty—the kind created by major infrastructure moves like this. If you want a practical starting point, begin with scenario modeling on your top 20 rail-exposed lanes and build an exception playbook that operations can run without improvising.

Where will your network be most brittle if rail choices narrow—and what would it take to make those lanes resilient before the next peak season hits?