Commerce Is Infrastructure: What a Decade of Big Tech Failures Proves About Owning Your Platform
From Meta to Google to OpenAI, every tech giant that tried to own the checkout eventually retreated. Commerce is infrastructure — not a feature you bolt onto a chat window or social feed. Own your platform, own your customer data, own your integrations, and build on a commerce engine built to last – Spree Commerce open source.
The biggest names in tech keep learning the same expensive lesson
OpenAI just “scaled back” Instant Checkout inside ChatGPT. After six months of positioning itself as the future of shopping, the company quietly moved purchases back to external retailer apps — back to the commerce platform.
But OpenAI isn’t the first tech giant to try owning the transaction. Over the past decade, Meta, Google, Twitter, Pinterest, Snapchat, TikTok, and now OpenAI have all attempted to turn their platforms into storefronts. Every single one has either retreated, scaled back, or quietly abandoned the effort.
The pattern is unmistakable. And the lesson is always the same: commerce is infrastructure, not a feature you bolt onto something else.
If you’re running a business that depends on ecommerce — whether that’s a direct-to-consumer brand, a marketplace connecting vendors and buyers, or a multi-brand portfolio serving customers across regions — this pattern should matter to you.
Because every time a sales channel captures your checkout, it takes away your customer, your data, and your ability to grow on your own terms.
Meta spent five years trying to own checkout — and gave it all back
Meta’s journey with native checkout on Facebook and Instagram is perhaps the most dramatic cautionary tale.
It started with big ambitions. Instagram Shopping launched with the promise that customers would never have to leave the app. Discover a product in your feed, tap, buy — all without opening a browser. Meta invested heavily, building out product catalogs, payment processing, and a full in-app checkout experience.
Then things got aggressive. In 2023, Meta announced that Shops without native checkout would be phased out. By April 2024, if you wanted to tag products on Instagram or Facebook, you had to use Meta’s built-in checkout. No exceptions.
And then came the reversal. In 2025, Meta abandoned native checkout entirely. All Facebook and Instagram Shops transitioned back to external website checkout. By September, the feature was gone.
What happened? By processing payments directly, Meta became the merchant of record — the entity on the hook for refunds, returns, chargebacks, fraud, and customer service for every transaction.
That’s the trade-off of owning the checkout: you get the customer data and the relationship, but you also inherit all the liability that comes with it. Running commerce means handling sales tax compliance across thousands of jurisdictions, fraud prevention, dispute resolution, and post-purchase support.
That’s a massive operational burden for a company built to sell ads, not manage shipping labels. Meta discovered what every commerce operator already knows: the transaction layer is the hardest part. It’s not a feature you tack on — it’s an entire business.
Along the way, Instagram also killed off live shopping in March 2023. The “Shop” tab disappeared. Creator commerce, visual search shopping, and community-driven shopping experiments — all shelved. Five years of investment, billions of dollars, and Meta ended up right back where it started: a discovery engine that sends customers somewhere else to buy.
For any merchant who built their checkout flow inside Instagram during that window, the disruption was real. Product catalogs needed migrating. Customer data was locked in a platform that changed its mind. Order history lived somewhere you didn’t control.
That’s the cost of building on someone else’s infrastructure.
Google tried three times and couldn’t make it stick
If Meta’s story is dramatic, Google’s is persistent. They’ve tried to own the buy button at least three times.
First came Google Shopping Express — a same-day delivery service launched in 2013 to compete with Amazon. Shut down in 2019. Then came Buy on Google, a native checkout feature that let people purchase directly from Google Search and Shopping results. Google killed it in September 2023. The Google Shopping app? Also shut down, in May 2021.
Each time, the pitch was the same: keep the customer inside Google’s ecosystem from search to purchase. Each time, the outcome was the same: merchants didn’t want to hand over their customer relationships, customers didn’t change their buying habits, and the operational complexity of running actual commerce infrastructure proved to be far more than a search engine wanted to manage.
Google is the most visited website on the planet. If anyone could have forced native checkout to work through sheer traffic volume, it would have been them. They couldn’t.
Each attempt required Google to take on merchant-of-record responsibilities — handling payments, managing refund liability, dealing with fraud. And each time, the operational cost of actually owning the transaction outweighed the strategic value of keeping users inside the Google ecosystem. Google wanted the data. It didn’t want the liability.
Twitter, Pinterest, and Snapchat all tried buy buttons — none of them worked
Remember when every social platform had a buy button? Around 2015–2016, it felt like the next big thing.
Twitter’s buy button was one of the first to launch and one of the first to die. The platform wasn’t built for commerce. Users came for conversation, not shopping. The button disappeared quietly, and Twitter never tried again.
Pinterest seemed like the perfect fit. People pin products they want to buy. The intent is right there. But even with 10,000 merchants joining the buyable pins program, conversion was abysmal. Major retailers reported fewer than 10 purchases per day through the feature. The gap between “I like this” and “I’ll buy this right here” turned out to be enormous.
Snapchat’s commerce features never gained traction either, with single-digit percentages of users ever making a purchase through the platform.
The common thread? Transactions were clunky, inventory was poorly synced, and consumer intent was fundamentally misunderstood. Browsing is not buying. Discovery is not checkout. And bolting a purchase flow onto a platform built for something else doesn’t change that.
TikTok Shop — spending billions to prove the rule
TikTok is the latest platform throwing enormous resources at owning commerce. TikTok Shop launched with the full weight of ByteDance behind it — subsidizing sellers, running massive promotional campaigns, building fulfillment infrastructure from scratch.
The results have been mixed. Sellers report a platform riddled with bugs, AI systems that erroneously remove products, and fulfillment services plagued by inventory errors and shipping delays.
When TikTok tried to force U.S. sellers into its own fulfillment network, the backlash was so severe the company had to pause the rollout.
Even when a platform is willing to spend billions, the operational complexity of commerce — inventory management, fulfillment logistics, tax compliance, returns processing, fraud prevention — doesn’t get easier just because you have a large user base. If anything, scale makes every gap more painful.
And by acting as the merchant of record, TikTok took on the full weight of customer liability — refunds, disputes, fraud — across a seller base it couldn’t effectively vet or monitor. That’s a liability profile that would be challenging for a dedicated commerce company. For a short-video platform, it’s existential.
Publishers tried to become stores — and found out they couldn’t
The platform graveyard extends beyond social media. In the mid-2010s, online publishers saw shoppable content as their salvation. If you’re already writing about the best running shoes, why not let readers buy right from the article?
BuzzFeed went all in. They built checkout technology, hired warehouse management specialists, and developed customer acquisition teams specifically for commerce. Commerce revenue grew initially — up 67% year over year from 2019 to 2020.
But it couldn’t save the business. Commerce revenues dropped as retailers pulled back. The news division shut down entirely in April 2023. Mashable tried the same playbook and fared even worse, eventually selling for a fifth of its peak valuation.
The publisher commerce story is a quieter version of the same lesson: knowing what people want to buy isn’t the same as being able to sell it to them. The infrastructure layer — payments, inventory, fulfillment, customer service, tax compliance — has to exist somewhere. Embedding shoppable widgets couldn’t replace it.
ChatGPT’s retreat is just the latest confirmation
Which brings us back to OpenAI. When ChatGPT launched Instant Checkout in September 2025, the pitch was compelling: ask an AI for a product recommendation, then buy it without leaving the conversation. Etsy stock jumped 16% on the announcement day. Shopify surged 8%.
The hype was enormous. But here’s what actually happened.
Out of millions of Shopify merchants, only about a dozen ever went live with Instant Checkout. Shopify’s dedicated ChatGPT integration landing page now quietly redirects to their homepage.
OpenAI’s own internal teams discovered that users were “high-intent browsers with low conversion rates” — people happily researched products in ChatGPT but wouldn’t pull the trigger on a purchase there.
According to Forrester’s December 2025 Consumer Pulse Survey, only 23% of GenX online adults had even used ChatGPT for product search. For older demographics, adoption was in the single digits.
The operational gaps were just as damning. As of February 2026, OpenAI still hadn’t built a system to collect and remit U.S. state sales taxes — a capability that Amazon, Shopify, and every serious commerce platform handles automatically across thousands of jurisdictions.
Real-time inventory synchronization across millions of retailers proved far messier than expected. Fraud detection infrastructure didn’t exist. And the planned 4% merchant fee sat on top of existing payment processing and platform fees, making the economics unworkable for most product categories.
Forrester principal analyst Emily Pfeiffer didn’t mince words: she said she was “shocked at the promises versus reality.”
The market’s reaction was swift and telling. The day the news broke, while the broader S&P 500 fell 0.5%, travel and commerce platform stocks surged in the opposite direction. Expedia jumped roughly 12%. Booking Holdings gained around 8%. Tripadvisor climbed 5%.
Mizuho, a leading Japanese investor, immediately bumped Booking to their top pick. Jefferies predicted Shopify could rally another 15% as customers and merchants flowed back through Shopify’s own checkout.
The market was essentially saying: the threat of AI disintermediating commerce platforms was overblown.
And here’s the deeper point that the stock market reaction revealed. When OpenAI tried to become the merchant of record — the entity processing the payment, owning the customer relationship, handling the liability — it took on a role it wasn’t built for.
Being the merchant of record means owning refunds, managing returns, bearing liability for fraud, and providing customer support. That’s a prize worth fighting for, because it means owning the customer data and the relationship that drives repeat business and upsell opportunities. But it’s also a massive operational burden that requires dedicated infrastructure.
OpenAI tried to replicate the full commerce stack without owning any of it. That’s what failed.
OpenAI’s retreat follows the exact same pattern as every example before it. A powerful platform with massive reach tries to capture the transaction. It works for discovery. It fails at checkout. The transaction goes back to the commerce platform.
Why commerce is infrastructure — and infrastructure demands a dedicated engine
So why does this keep happening? Why do the most powerful, best-funded technology companies in the world keep failing at the same thing?
Commerce isn’t a feature. It’s a full-stack operation.
Start with the simplest case: a single product purchased from a single brand. That’s what a buy button on Instagram, a TikTok Shop listing, or ChatGPT’s Instant Checkout was trying to enable.
Even this “simple” transaction requires a product catalog with accurate pricing, real-time inventory checks, payment processing with PCI compliance, sales tax calculation across thousands of jurisdictions, shipping rate calculation, order confirmation, fulfillment coordination, and post-purchase support for returns and refunds. Miss any one of these and the customer experience breaks.
But here’s the part that matters most, and the part every big tech company underestimated: whoever processes the payment becomes the merchant of record. That means owning the customer relationship and the data that comes with it — the ultimate prize that drives repeat business and upsell opportunities.
But it also means bearing liability for refunds, returns, chargebacks, fraud, and customer service issues. At scale, this operational burden isn’t just significant — it’s a full-time business.
It’s why Amazon, the company that actually succeeded at platform commerce, spent decades and billions building dedicated infrastructure for exactly this: fulfillment centers, fraud detection systems, customer service operations, seller performance monitoring, tax compliance engines.
Commerce was never Amazon’s side project. It was the entire company.
Now scale up from single-vendor to multi-vendor. This is what happens when a platform allows marketplace-style shopping — multiple products from different sellers purchased in a single order, like Amazon. Suddenly everything gets harder.
Orders need to split across vendors. Payments need splitting too — the platform takes its commission, each vendor gets their payout, and the timing and currency of those payouts need managing. Refunds become multi-party problems: if a customer returns one item from a three-vendor order, the system needs to calculate the right refund amount, claw back the right commission, and process the right payout adjustment for the right vendor.
Each vendor ships separately, often with different shipping costs, different carriers, and different delivery windows. The platform needs to monitor vendor performance, manage disputes between buyers and sellers, and coordinate all of it without losing track of a single transaction.
That’s what a multi-vendor marketplace actually requires — and it’s why every channel that tried to own the checkout gave up once they hit this layer of complexity.
No social media platform is going to build that. No chat interface is going to manage that. No publisher’s shoppable widget is going to handle that.
Commerce platforms have spent years — in some cases, decades — building this infrastructure. It’s not something you replicate by adding a checkout button to a chat interface, a social feed, or an article page.
Own your platform, own your customer, own your data
The lesson from a decade of big tech failures isn’t that AI or social commerce are dead. AI is a brilliant discovery engine. Social platforms drive incredible product awareness. Publishers create genuine purchase intent.
But none of them should own your checkout. Here’s why.
Every time a platform captures the transaction, it becomes the merchant of record — and it captures the customer along with it. It controls the data. It sets the fees. It decides the terms. And when that platform changes strategy — as every single one eventually does — you’re left rebuilding from scratch.
That Instagram product catalog you spent six months building? Gone when Meta changed course. Those Google Shopping listings you optimized? Worthless when Buy on Google shut down. The customer data from those transactions? Locked in a platform that changed its mind.
Being the merchant of record is the ultimate prize in commerce — it’s how you build the customer relationships that drive repeat business, lifetime value, and upsell opportunities. But it comes with real responsibilities: refund liability, fraud risk, tax compliance, and customer service.
That’s exactly why big tech keeps retreating from it. They want the data and the revenue share, but not the operational burden. The companies that thrive in commerce — the ones that actually earn customer loyalty — are the ones willing to own that full stack.
Your commerce engine should be the source of truth — the one thing that doesn’t change when a sales channel pivots. It should handle the transactional logic, guarantee customer and order data security, enforce compliance, and serve as the central point for every integration your business depends on: marketing automation, payments, taxation, shipping, and post-purchase care like returns, refunds, and dispute resolution.
With an API-first, open-source eCommerce platform, that’s exactly what you get. Your commerce engine powers shopping experiences across every sales channel — web storefronts, mobile apps, B2B portals, multi-vendor marketplaces, multi-tenant platforms, and yes, even AI-powered checkout. But regardless of where the customer discovers your product, the engine remains the backbone — handling inventory, processing orders, protecting data, and keeping every integration connected.
That’s what separates infrastructure from features. Meta turned checkout on, then off. Google built buy buttons three times and removed them three times. OpenAI’s Instant Checkout lasted less than six months. But the commerce engine underneath? That keeps running.
Commerce is too important to leave in the hands of sales channel owners who will inevitably change strategy, pivot priorities, or shut down features when the economics don’t work out. Be the merchant of record.
Own the customer relationship and the data that comes with it. Own your platform. Own your integrations. Own your security and compliance. Build on commerce infrastructure that lasts.
Get started with an open-source commerce engine you control, or book a call to explore how it can power commerce across every channel your business needs — without giving up ownership of anything.
