Amazon Business hits $60B and Instacart bets on computer vision as B2B ecommerce enters its next infrastructure cycle

Amazon Business crossed $60 billion in annualized gross sales, Digital Commerce 360 reported on July 21, the same day Instacart disclosed its acquisition of Arpalus, a computer vision company that automates in-store shelf intelligence for grocery retailers. The back-to-back announcements are not coincidental in their timing. They reflect a single, accelerating dynamic: the physical and digital layers of commerce are converging, and the enterprise infrastructure connecting them is being rebuilt at speed.

Amazon Business at $60 billion: a procurement channel that demands a strategy

The $60 billion annualized gross sales figure, reported by Digital Commerce 360 citing Amazon Business press materials, cements the platform's standing as one of the largest B2B ecommerce destinations in operation. For procurement directors who have historically treated Amazon Business as a convenience channel for office supplies and incidental spend, the scale now warrants a more deliberate posture. The platform has grown well beyond tail spend into categories including industrial, laboratory, and facility supplies, with business-specific pricing, analytics, and purchasing workflows.

Forbes, in its updated ecommerce statistics report audited through July 2026, notes that B2B ecommerce overall continues to outpace B2C growth in dollar volume, driven by the digitization of procurement workflows that were manual as recently as five years ago. Amazon Business is the clearest single proof point: its buyer base spans small businesses to enterprise accounts and government entities, and the platform's contract pricing and approval-routing features are designed specifically for multi-stakeholder procurement environments.

The operational implication for sourcing teams is catalog visibility. Suppliers whose product data is incomplete, unstructured, or missing key attributes are effectively invisible to Amazon Business's search and recommendation systems. That problem is about to compound significantly.

Instacart and Arpalus: computer vision closes the physical data gap

Instacart's acquisition of Arpalus, reported by Mary Meisenzahl at Digital Commerce 360, brings a specific capability into the grocery fulfillment stack: computer vision that can read physical shelves, identify out-of-stock conditions, detect planogram compliance failures, and generate real-time inventory signals without manual scanning. For grocery retailers and their CPG suppliers, that matters because the biggest friction in omnichannel fulfillment has always been the lag between what is actually on the shelf and what the digital ordering system believes is there.

Arpalus's technology sits at the point of truth: the physical shelf. By integrating that data stream into Instacart's platform, the acquisition creates a tighter feedback loop between store-level inventory reality and the consumer-facing or B2B ordering interface. Retailers using Instacart's fulfillment infrastructure stand to see fewer substitution events and fewer customer-facing out-of-stock errors, both of which carry direct revenue and satisfaction costs. For CPG procurement and supply chain teams, the downstream effect is more precise replenishment triggers based on what is actually depleted rather than what a periodic audit estimated.

The move also signals where grocery technology investment is concentrating. Rather than building computer vision in-house, Instacart chose acquisition, suggesting the category is mature enough that buying proven capability is faster than developing it. Operators evaluating their own shelf-intelligence or store-execution technology vendors should take note of the validation this deal provides for the category.

Agentic AI adds a third pressure point on catalog and discovery

Alongside the Amazon Business and Instacart news, Digital Commerce 360 published guidance from Deloitte on preparing B2B ecommerce infrastructure for agentic AI discovery. Agentic AI refers to autonomous software agents capable of executing multi-step tasks, including product research, vendor comparison, and purchase initiation, without direct human instruction at each step. The practical implication for a B2B ecommerce operator is that the buyer reviewing a catalog page may increasingly be a machine, not a person.

The Deloitte analysis cited by Digital Commerce 360 focuses on data readiness as the foundational requirement. Product attributes, pricing structures, availability signals, and contract terms all need to be machine-readable and consistently structured for an AI agent to evaluate and act on them correctly. Organizations that have invested in clean, well-governed product information management systems are better positioned than those still relying on PDF catalogs or inconsistently tagged SKU data.

This is not a distant concern. Several enterprise procurement platforms are already piloting agent-assisted purchasing for routine, high-frequency categories. The companies that will be found and selected by those agents are the ones whose product data is structured to answer machine queries, not just human keyword searches. According to Forbes, ecommerce as a share of total retail and B2B purchasing continues to climb in 2026, which amplifies both the opportunity and the risk of being de-listed or deprioritized by algorithmic buyers.

What this means for your team

• Audit your Amazon Business catalog now: at $60 billion in annualized gross sales, completeness and accuracy of product attributes directly determine visibility to a massive buyer pool. Treat it as a first-party channel, not a spillover.
• Evaluate shelf-intelligence and store-execution vendors with Instacart's Arpalus acquisition as a benchmark: the deal validates computer vision for in-store inventory as a mature, integrable capability, not an experimental one.
• Assess your product data architecture for machine readability: agentic AI buyers evaluate structured data, not marketing copy. Prioritize clean attribute sets, consistent taxonomy, and API-accessible catalog feeds over SEO-optimized descriptions written for human readers.
• Map your replenishment triggers to digital fulfillment signals: as platforms like Instacart close the loop between physical shelf data and digital ordering, suppliers who can respond to real-time depletion signals rather than periodic purchase orders will hold a service-level advantage.