Adding a Second Warehouse to an Amazon-Heavy Network: When Splitting Inventory Pays and When It Doesn't

By Bernardo Campelo, Forbes Business Council E-Commerce Leader, Amazon SPN Certified provider, Amazon SP-API authorized partner, and Founder of PrepVia.

As of October 29, 2026, the Amazon network your inventory feeds is regional, not national. Amazon says it moved its US fulfillment network "from a national fulfillment network to a regionalized network model," built as "eight interconnected regions in smaller geographic areas," each able "to operate in a largely self-sufficient way, while still being able to ship nationally when necessary." That is how Amazon describes it in its 2022 letter to shareholders. So when a brand that sells mostly on Amazon asks whether it needs a second warehouse, part of the answer is already settled: for FBA orders, Amazon places the stock close to customers itself.

The second warehouse question lives everywhere else: direct-to-consumer orders, wholesale accounts, other marketplaces, and stock waiting to go into Amazon. A second node can cut transit time on those orders, and it costs money in places most models miss: a second safety stock, a weaker AWD discount, and a second count that has to agree with the first.

Every date and name below was checked on October 29, 2026 against Amazon's About Amazon pages, Seller Central help, sell.amazon.com, the USPS Domestic Zone Chart and the FedEx 2026 Service Guide, linked where they first appear. Amazon revises fees and help pages, and carriers update zone charts, so open the live page before you model a number.

The Network You Are Adding a Node To

Before regionalization, Amazon says, "If a local fulfillment center didn't have the product a customer ordered, we'd end up shipping it from other parts of the country, costing us more and increasing delivery times." The fix was to keep a broad selection in each region and ship across regions only when necessary.

In July 2023, Amazon said "more than 76% of the orders we fulfill come from within the customer's region." In its 2023 letter, it credited regionalization, "where we re-architected the network to store items closer to customers," for faster delivery, and said its US cost to serve fell by more than $0.45 per unit in 2023. That is Amazon's cost, not yours. The pages I read on October 29, 2026 give no newer figure for in-region orders.

Amazon's inbound fees already follow regions. On the FBA inbound placement service fee page, a minimal split sends inventory "generally to a single location, for a fee," and Amazon decides how to place it across the network. The fee "may vary by inbound location," with higher fees possible for locations in the West. An Amazon-optimized split to multiple locations carries no fee, and to qualify, "shipments must include at least five identical cartons or pallets per item." The full fee math is in our 2026 inbound cost per unit model.

So for FBA demand, a second 3PL does not move units closer to customers; Amazon does that. It changes where your inbound starts, how far your own channels' orders travel, and how much stock you hold outside Amazon.

Four Ways to Add Capacity: One 3PL, Two 3PLs, AWD as the Second Node, or Multi-Channel Fulfillment

Most plans compare two options, one 3PL or two. Amazon sells a third. Its September 24, 2026 announcement says sellers "wanted one inventory pool to fulfill orders from every channel," and names Multichannel Fulfillment (MCF), Amazon Warehousing and Distribution (AWD) and Buy with Prime as the services it built for that.

AWD is the piece that works like a second node. Amazon's AWD guide says you can distribute "to the Amazon FBA fulfillment network" and also "to non-Amazon sales channels, including your own or 3PL warehouses, wholesalers, distributors, and other channels in the US like Walmart Fulfillment Services." That route is Multi-Channel Distribution (MCD).

For parcel orders from your own site or other marketplaces, run one more comparison before you sign a West 3PL. Amazon's announcement shows a seller who can "select a Shopify order, click to fulfill through Amazon," which ships the order from FBA inventory that Amazon already spreads across its regions. Price that route on Amazon's current Multi-Channel Fulfillment fees next to the cost of a second node.

Amazon's Multi-Channel Distribution fees page says the outbound transport fee "is distance-based and determined based on shipping weight and the shipping method you select." Effective January 15, 2026, box loads run from $0.50 per pound at 0 to 150 miles up to $1.26 per pound at 1,801 miles or more, and palletized loads from $0.11 to $0.42 per pound. MCD orders also pay AWD storage and processing.

Two more AWD limits matter. Since July 31, 2026, the AWD eligibility page says AWD only accepts sortable items, smaller than 18 x 14 x 8 inches and under 20 lb, so oversize stock needs a node outside AWD; our AWD size limit guide covers where it goes. And on timing, Amazon's AWD replenishment page says "the replenishments take up to an average of 14 days to move from an AWD center to the FBA network," and tells sellers to "expect delays in processing during the peak period."

The full AWD fee table is in Amazon AWD vs a 3PL in 2026; what matters here is that AWD pricing is regional too. Amazon's AWD fees page says that effective January 15, 2026 it applies "differential storage pricing between West Coast, East Coast, South East, and South Central regions." Base storage is $0.57 per cubic foot per month in the West Coast region and $0.48 in the other three.

What Carrier Zones Say About Where a Second Node Belongs

For orders you ship yourself, carriers price ground parcels by zone, and the zone comes from the origin and destination ZIP codes. The USPS Domestic Zone Chart states it plainly: "The first 3-digits of your destination ZIP Code determine the zone." I pulled it on October 29, 2026 for shipping dates in October and in November 2026, editions effective October 1 and November 1, 2026, and the zones below are the same in both.

Here is that chart for four candidate origins. These are USPS zones only. UPS and FedEx publish their own charts keyed to your origin ZIP; the FedEx 2026 Service Guide tells shippers to "enter your origin ZIP code and destination ZIP code," prices contiguous US ground in Zones 2 through 8, and warns that "ZIP code and zone alignments are subject to change without notice."

A Miami origin is Zone 8 to Los Angeles, San Francisco, Seattle and Phoenix, and Zone 6 to New York, Philadelphia, Boston, Chicago and Dallas. No Southeast warehouse is a national parcel node. The table supports a pairing: an East or Central node, plus a West node when your own-channel demand there is large and steady enough to carry a second stock.

The zone test does not apply to FBA stock, which reaches the customer from whichever Amazon region holds it. For the Amazon share of a catalog, the question is inbound distance to the fulfillment centers you are sent to, plus the placement option you pick.

The Safety Stock You Pay For Twice

The cost of a second node that most budgets leave out is the extra buffer. Safety stock follows the variability of demand over the replenishment lead time. When one demand stream is split across two warehouses, each warehouse has to cover its own swings, and the swings no longer cancel out inside one pool.

Our guide to where shared Amazon and Walmart stock should sit sets out the textbook formula. Under its assumptions, an even split needs about 41% more total safety stock at the same service level, because √2 ≈ 1.41. Brands rarely split evenly, so here is the same arithmetic for uneven splits.

This is our illustration, not a figure from Amazon or a carrier. It assumes the demand at each node is independent, the demand variance at each node is proportional to its share, and the two nodes have the same lead time and service level. Each node's swings then scale with the square root of its share, so the total safety stock multiplier is √p + √(1 − p).

Two readings matter. A lopsided split does not remove the penalty: 10% of demand in a second node still adds about a quarter more safety stock under these assumptions. And the penalty shrinks when demand at the two nodes moves together. Large swings and an unknown regional mix enlarge each node's buffer, which is why a second node suits SKUs with stable regional demand better than new or promotional SKUs.

The extra buffer also pays storage at two sites. If a SKU's parcel savings from lower zones do not cover its second buffer and that storage, the SKU stays in one node.

The 70 Percent Rule and the Direct Inbound Trap

If your Amazon flow runs through AWD, a second warehouse can cost you fee exemptions and storage discounts without any extra fee showing up in the model. Amazon's replenishment page defines the ratio as "Auto Replenishment shipment quantity / (Auto + Manual + Direct Inbound to FBA) = Auto Replenishment Ratio." It is "calculated on a daily basis and takes the quantities shipped in the past 90 days only."

At 70% or more, that page says the ASIN "will be exempt from low inventory level and storage utilization fees." The AWD fees page applies the same test per SKU and adds the aged inventory surcharge for products stored between 181 and 365 days. The same 70%, with at least 70 days of combined supply, is also required for AWD's discounted storage rates: smart storage, 10% off base, and Amazon Managed, 20% off. Amazon's own example: 7,000 auto-replenished units, 1,000 direct inbound and 1,500 manual replenishment give 7,000 / 9,500 = 73.6%.

The trap follows from the formula. A second warehouse that ships a SKU straight into FBA, for speed or to hit a deadline, adds to the denominator only. Take Amazon's example and add 1,000 more direct-inbound units from the new node: 7,000 / 10,500 = 66.7%, and the SKU drops below 70%. The weekly protocol for watching this is in the 70 percent rule. For network design, the rule is simple: decide per SKU whether FBA is fed by AWD or by your warehouses, and do not let a second node become a third feed by accident.

Who Holds the Source of Truth

With one 3PL, the warehouse count and the Amazon count are two numbers to reconcile. With two 3PLs, AWD and FBA, they are four, plus everything in transit between them. None of Amazon's reports can be the master record for that network, because each one sees only part of it.

• FBA. Amazon's Inventory Ledger report "is like a bank statement for your inventory," with up to 18 months of history, and covers units in Amazon fulfillment centers, including "units that are in transit between fulfillment centers."
• AWD. The AWD Inventory Ledger report is a separate report about "events which affect inventory levels in AWD distribution centers."
• AWD and FBA together. Amazon's combined days-of-supply metric on the AWD fees page excludes "units that are currently being inbounded to AWD or directly to the Amazon fulfillment network."
• Your warehouses. Stock at your 3PLs appears in none of these reports.

Transit time is where counts drift. Amazon's shipment tracking page says check-in typically takes 2 to 5 days after delivery, up to one more day to reach Receiving, and 6 to 15 days for units to become Prime eligible after the first Amazon facility receives them. And Amazon notes that "AWD to FBA shipments are generated internally, and there is no POD or BOL generated for replenishment shipments."

So the master record has to sit above every node: your ERP or order management system, holding each warehouse, AWD, FBA and every transfer, with a status for stock between nodes. Amazon's Seller Central multichannel tools are a different layer. The September 24, 2026 announcement says they are rolling out "gradually to sellers in the U.S. store over the coming months." They bring orders, listings and profitability into one view, and Amazon says sellers can already "fulfill across channels in a few clicks," sending an off-Amazon order to Amazon for fulfillment. The announcement does not describe them as moving or counting stock across your own warehouses.

The month-end routine for matching the counts, with the five reasons they disagree, is in three-way inventory reconciliation. Before signing a second 3PL, settle three things in writing:

• The transfer rule. Who creates a transfer, and when each node deducts or adds it.
• The adjustment rule. Who posts a count adjustment, with what evidence, and how fast.
• The in-prep status. Units being labeled or kitted are neither sellable nor gone, so they need their own status.

When Splitting Inventory Pays and When It Doesn't

Decide SKU by SKU. A second node tends to pay when:

• The demand is yours to ship. Direct-to-consumer, wholesale or marketplace orders that Amazon's regions do not fulfill for you, with a large share far from your first node.
• The regional demand is steady. Proven sellers whose regional split has held for several quarters, so the second buffer stays small.
• The zone gap is real on your own chart. Your carriers' charts show a large zone drop for that demand, and the parcel savings cover the second buffer and its storage.
• The SKU cannot use AWD. Oversize items that AWD's sortable limit excludes have to be stored outside AWD, so their reserve stock sits with a 3PL either way. That argues for a second node only when the oversize demand you ship yourself is far from the first one.

A second node tends not to pay when:

• The SKU is mostly FBA. Amazon already spreads that stock across its regions after receiving, and a second 3PL only adds an inbound origin.
• The SKU runs on AWD auto-replenishment. Direct inbound from a new node can push it under the 70% ratio.
• Demand is new, seasonal or promotional. Large swings make the split buffer expensive, and the regional mix is unknown.
• You cannot yet track transfers. If your system of record cannot hold node-to-node transfers and in-prep stock, a second node adds unreconciled units.

"Splitting inventory is not a goal. It is a cost you accept when a channel's customers sit too far from the warehouse that serves them. Put each flow where it performs best, keep one record of every unit, and split only the SKUs whose regional demand is steady enough to carry the extra safety stock."

Bernardo Campelo, Founder of PrepVia

The change to plan for in 2027 is narrow: Amazon's multichannel tools make it easier to sell one SKU on more channels and to send off-Amazon orders to Multi-Channel Fulfillment, but they do not move or count units in your own warehouses. If your channel mix shifts next year, rerun this test against all four options before you sign a second contract. For contracting a second node, our guide to putting Amazon Supply Chain Services in a 3PL RFP shows how to compare bids line by line.

Where a Prep Center Fits

A network plan should name which kind of node it needs. A prep center is a pass-through that receives, labels and ships inventory into Amazon; a fulfillment center holds stock and ships to consumers (the difference, explained). A West node for direct-to-consumer orders is a fulfillment center decision.

PrepVia is an Amazon FBA prep center at 8507 NW 72nd St, Miami, FL 33166, and it is Amazon SPN Certified. It receives inventory, including imported freight, preps it to Amazon's requirements and ships it to Amazon FBA or AWD, using Amazon partnered LTL and FTL. Our standard prep window is 24-36 hours. Discrepancies are reported within 24 hours of arrival, with photos, and the app tracks expected versus received units in real time. PrepVia also preps to Walmart WFS and TikTok Shop rules. That is prep, not an API integration with those channels.

The honest role in a two-node network is the Amazon and AWD inbound node for freight that lands in South Florida. Shared FastLane is one example: a full Amazon Freight truckload PrepVia books for several clients going to the same Amazon fulfillment center, with freight split per pallet, per client. If your direct-to-consumer volume needs a West or Central node, that is a different warehouse, and we say so. EDI and ASN transactions are not offered, so a network that runs on EDI between nodes needs that from another provider.

Ask any provider you add the three source-of-truth questions above. Our page for ecommerce brands shows how PrepVia works with brands that sell on more than Amazon. Pricing is quoted per account on a free 15-minute call.

Frequently Asked Questions

When does a second 3PL warehouse pay for itself?

When the orders you ship yourself have steady demand far from your first warehouse, and the parcel savings from lower zones cover a second safety stock and its storage. For stock that goes into FBA, Amazon already places units across its regional network after receiving, so a second 3PL mainly changes where the inbound starts. Price Amazon's Multi-Channel Fulfillment for those off-Amazon orders before you add a node.

How much extra safety stock does a second warehouse need?

Under textbook assumptions (independent demand at each node, the same lead time and service level), an even split needs about 41% more total safety stock than one pool. A 70/30 split needs about 38% more and a 90/10 split about 26% more. This is our illustration, not an Amazon figure. The penalty shrinks when demand at the two nodes moves together.

Can AWD work as my second warehouse?

For some SKUs. Amazon says AWD can distribute to the FBA network and, through Multi-Channel Distribution, to your own or 3PL warehouses, wholesalers, distributors and Walmart Fulfillment Services. MCD transport is priced by distance and weight. Since July 31, 2026, AWD accepts only sortable items, and Amazon puts AWD-to-FBA moves at up to an average of 14 days, with delays in peak.

Does a second warehouse affect the AWD 70% rule?

It can. Amazon's ratio is auto-replenished units divided by auto, manual and direct-to-FBA inbound units over the past 90 days. Units a second warehouse ships straight into FBA count only in the denominator. Below 70%, the SKU loses the exemption from the low-inventory-level fee, the storage utilization surcharge and the aged inventory surcharge for units stored 181 to 365 days, plus AWD's 10% smart storage or 20% Amazon Managed storage discount.

Is a Miami or Southeast warehouse a national node?

Not for parcels. On the USPS zone chart for fall 2026 shipping dates, a Miami origin is Zone 8 to Los Angeles, San Francisco, Seattle and Phoenix and Zone 6 to New York and Chicago. A Southeast node fits Amazon and AWD inbound and Southeast parcel demand. Check your own UPS and FedEx charts too.

What should be the source of truth with several stock locations?

Your own ERP or order management system. Amazon's Inventory Ledger covers its fulfillment centers, AWD has a separate ledger, and neither sees your 3PLs. Your system has to hold every node, every transfer, and stock in transit or in prep.

Final Take

A second warehouse is a tool for the demand Amazon does not ship for you. For FBA stock, Amazon's regions already put units near customers, and a second 3PL mostly adds a second buffer, a second count and a risk to the AWD ratio. For direct-to-consumer and wholesale demand that sits far from your first node, the zone chart can justify it, SKU by SKU, when the regional demand is steady.

Plan 2027 with four columns: one 3PL, two 3PLs, one 3PL plus AWD, and Multi-Channel Fulfillment for off-Amazon parcels. Price each with your own zone charts and Amazon's current fee pages, and keep one system of record above all of them.