Earlier this week, parts of TikTok were knocked offline due to an Oracle cloud outage.
This marked the second time in just over a month that the vendor caused the social media behemoth to go dark.
Oracle Cloud Infrastructure’s US East (Ashburn) region began experiencing connection timeouts and elevated latency on the afternoon of the 3rd of March, with the root cause not identified until just after midnight.
This means that some US TikTok users had been dealing with a degraded or completely broken experience for the better part of 10 hours.
The TikTok USDS Joint Venture – formed following TikTok’s US ownership restructuring, in which Oracle holds a 15% stake – posted a statement on X:
“An issue with an Oracle data center is impacting some parts of the TikTok U.S. user experience. Creators may temporarily experience lags in posting content while Oracle works to resolve the issue.”
The response from users was less diplomatic. Replying to Oracle’s assurance that it was “working quickly to restore normal service operation,” one user fired back: “do y’all ever plan on getting it back up?? this is ridiculous. 10hrs is not quickly restoring services.”
For the CX industry, the more pressing takeaway is less about TikTok specifically, and more about what happens when a customer-facing platform collapses because of a single cloud provider.
A Pattern That’s Becoming Difficult to Explain Away
As noted above, this was not a one-off. A winter storm took out another Oracle data center in early February, which again impacted TikTok.
Two significant outages in five weeks make Oracle CEO Larry Ellison’s 2022 claim that the company’s cloud “doesn’t go down” increasingly hard to defend.
To be fair to Oracle, it isn’t the only cloud provider that’s had to answer for reliability failures recently.
AWS and Azure both experienced notable regional outages in late 2025.
In response to these outages, Gartner Distinguished VP Analyst Lydia Leong wrote in an article titled, Don’t Let Outages Erode Your Trust in the Cloud:
“No cloud provider can promise zero downtime. What sets IT leaders apart isn’t whether they avoid incidents altogether; it’s how they prepare and respond.”
That message carries more weight given Forrester’s recent forecasting. In the company’s Predictions 2026: Cloud Computing Report, the analyst firm predicted that at least two major multiday cloud outages will occur this year.
The reason, according to Forrester, is that hyperscalers are diverting infrastructure investment toward GPU and AI workloads, leaving aging x86 environments increasingly exposed to failure under growing complexity.
Forrester also noted that customers who have already weathered recovery challenges will start pushing cloud providers harder on operational risk – well beyond the current focus on AI service adoption.
The CX Stack Is Not a Special Case
For contact center and CX technology leaders, the Oracle-TikTok situation is a useful, if uncomfortable, reference point.
Most modern CX stacks – cloud contact center platforms, digital engagement tools, AI-powered agent assistance, omnichannel routing – run on top of one or two major cloud providers. When those providers have a bad day, the downstream effects land directly on customers.
Most customers don’t view 10 hours of degraded service as an infrastructure problem; they see only a broken experience with the brand.
The reflex response when something like this happens is often to pivot toward multicloud.
Gartner pushes back on that, however, arguing that “pursuing multicloud resilience can cost more than it saves, introducing technical complexity without truly eliminating systemic risk.”
Instead, the firm recommends building out resilience within a single cloud first, through multi-availability-zone architecture, regular failover drills, and what Gartner calls “application substitutability”: having backup platforms or manual workarounds ready for any function where extended downtime is simply not an option.
The Questions CX Leaders Should Be Asking
The Oracle incidents raise some practical questions worth revisiting now rather than after the fact.
Do you know which cloud or clouds your CX vendors are actually running on? Do your contracts include meaningful SLAs with real consequences, or just reassuring language? And if your primary platform went dark for 10 hours, what would your contact center do in the meantime?
Gartner also highlighted transparency as a meaningful differentiator between providers, noting that those who are open about their infrastructure dependencies give CIOs “actionable data for risk management instead of leaving them guessing.”
That’s worth weighing in vendor evaluations, particularly as the CX technology market consolidates around a smaller number of major platforms and the cloud infrastructure sitting underneath them.
Oracle confirmed the issue was resolved at 09:24 UTC on March 4. The Register contacted the company for an explanation and reliability assurances; at the time of writing, no response had been issued.
