Google made its move on Canva’s home turf on September 1, launching Pics, a Workspace-native AI image editor, into general availability just three weeks after Canva’s own investors slashed billions off the design company’s valuation. The timing is not a coincidence industry watchers are ignoring. Canva spent the back half of 2026 explaining to shareholders why its AI features cost more to run than expected. Google spent the same stretch quietly building the exact kind of AI-first design tool that made those costs a problem in the first place.
According to Startup Daily and TheNextWeb, Canva’s earliest and largest venture backers, Blackbird Ventures and AirTree Ventures, cut their internal valuation estimate for the company by 17%, from roughly US$42 billion to US$34.9 billion, after commissioning independent outside assessments. The Australian Financial Review first reported the markdown, and Capital Brief reported that some offshore investors believe the real number could be even lower. Canva’s own internal valuation, the one used to price employee share issuances, dropped further still, from US$38.9 billion to US$31 billion over the past year, per the same reporting.
Then, on September 1, Google pushed Pics to general availability across Workspace and its consumer AI subscription tiers, positioning it directly against the product category Canva has owned for more than a decade. TechCrunch described it as “Google’s answer to Canva,” built around prompting rather than manual layout. For a company that just told Wall Street-adjacent investors that AI was eating its margins, watching the world’s biggest AI lab ship a free-to-low-cost rival inside tools that over three billion people already use is not a great way to start the fall quarter.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What actually happened: two stories colliding in three weeks
Strip away the coincidence of timing and there are two separate, well-documented events here. The first is financial: Canva’s AI ambitions got expensive faster than its board wanted. The second is competitive: Google shipped a product that competes directly with the thing Canva makes money from. Put together, they read like a cautionary tale about what happens when a company tries to become an AI platform while the actual AI labs are building the same features into products people already have open all day.
Canva’s valuation cut traces back to a shareholder update covered by Fortune in mid-August, in which CEO and co-founder Melanie Perkins told investors the company was lowering its 2026 revenue growth forecast from 30% to 20%. The stated reason was cost, not demand. Perkins said the “average cost of serving an AI task was too high” and that Canva had been “relying too heavily on frontier models” from outside vendors to power features it had rolled out aggressively across its Visual Suite. Rather than keep shipping new AI tools while the unit economics were underwater, Canva chose to slow the pace of new AI launches and pour engineering effort into cutting the cost per task.
That decision produced a genuine improvement: Canva says it has cut AI servicing costs by roughly 90% since the crunch began, largely by building proprietary in-house image and video models and by folding in the team and technology from its 2025 acquisition of Leonardo.AI. Blackbird partner Rick Baker, cited in the Forbes Australia report, pointed to the same cost reduction as evidence the company had gotten its arms around the problem. Canva has said its proprietary video model now runs roughly 17 times cheaper than comparable frontier models, and its image model roughly 30 times cheaper, though those are company-reported figures rather than independently audited ones.
None of that stopped the markdown. Blackbird and AirTree still moved first, cutting their valuation mark by 17% based on outside assessments. Canva’s own valuer, used for internal employee-equity pricing, cut its number by even more over the trailing twelve months. The SmartCompany writeup on the episode also connected the dots to Canva’s IPO plans: Blackbird had previously told its own investors the company was “ready” for a second-half 2026 US listing. That timeline is now widely expected to slip into 2027, since taking a company public on a valuation trajectory pointing down rather than up is not how founders like to time a debut.
Google Pics: what it is and why it lands where it does
Google Pics is not a new lab experiment. It reached general availability on September 1, 2026, according to Android Authority, after a more limited preview earlier in the year. It is built on Google’s Nano Banana image model (the second-generation version, sometimes referred to as Nano Banana 2), the same underlying technology Google has been rolling into Gemini, Workspace, and its consumer AI subscriptions throughout 2026.
Functionally, Pics behaves less like a template library and more like a conversational image editor. Per TechCrunch’s reporting, a user opens it either as an overlay directly inside a Google Docs or Slides file when they click on an image, or as a standalone tool at pics.new. From there, a written prompt generates an image, and follow-up prompts let the user retouch a single object, swap a color, or translate embedded text into other languages, all without regenerating or manually redoing the rest of the composition. The model is reported to recognize individual elements inside a picture, whether that picture was AI-generated or a real photograph, and treat each one as an editable layer, so a user can move, resize, or delete one object while the background stays untouched.
Access is tiered through Google’s existing subscription and enterprise structure rather than sold as a standalone product: Google AI Pro and Ultra subscribers get it, as do Google AI Pro for Education users, and Workspace customers on Business Standard, Business Plus, Enterprise Standard, and Enterprise Plus plans. That distribution strategy is the part that should worry Canva’s board more than the product itself. Canva has to convince people to sign up for a new tool. Google is dropping Pics into inboxes and slide decks people already have open.
Canva vs. Google Pics: how the two actually differ
The two products are not identical, and the differences matter for anyone deciding where to put design work in the fourth quarter of 2026. Canva remains a template-first, drag-and-drop platform with an enormous library of pre-built layouts, brand kit tools, a print business, and now a “Canva AI 2.0” conversational layer layered on top of that existing manual toolkit. Google Pics skips the manual layer almost entirely and starts from a prompt, using object-level segmentation to let a user refine specific pieces of an image after the fact.
| Category | Canva (Visual Suite + Canva AI 2.0) | Google Pics |
|---|---|---|
| Core interaction model | Template-based, drag-and-drop, with conversational AI layered on top | Prompt-first generation with object-level editing |
| Underlying AI model | Proprietary Canva Design Model, plus in-house image/video models built after the 2025 Leonardo.AI acquisition | Google’s Nano Banana (Nano Banana 2) image model |
| Access point | Standalone app and website, canva.com | Overlay inside Google Docs/Slides, or standalone at pics.new |
| Who can use it | Free and paid tiers open to any signup | Google AI Pro/Ultra subscribers, Google AI Pro for Education, and Workspace Business/Enterprise plans |
| Distinct strength | Template depth, brand kits, print fulfillment, existing user base | In-document editing, object segmentation, in-image text translation |
| Business model exposure | Subscription revenue tied directly to AI feature costs | AI cost bundled into existing Workspace/Google AI subscriptions |
The most consequential difference may be the last row. Canva sells design software, so every dollar it spends running an AI task is a direct cost against a product whose price it has to justify on its own. Google bundles Pics into subscriptions and enterprise contracts that already generate revenue from search, cloud, and productivity tools, which means the AI cost of Pics doesn’t have to clear the same bar on its own. That is precisely the dynamic Canva’s own shareholder letter was describing when it flagged the cost of serving AI tasks as the reason for slowing its roadmap.
The valuation math, in plain numbers
Canva peaked at a private valuation near US$42 billion in prior funding rounds before the current reset. The August markdown pulled the number down in stages, depending on which internal metric is used.
| Metric | Prior mark | Current mark | Change | Source |
|---|---|---|---|---|
| Blackbird/AirTree investor valuation | ~US$42 billion | ~US$34.9 billion | -17% | Startup Daily, TheNextWeb |
| Canva internal employee-equity valuation | US$38.9 billion | US$31 billion | Down over trailing 12 months | Capital Brief |
| 2026 revenue growth forecast | 30% | 20% | -10 percentage points | Fortune |
| Reported AI cost-per-task reduction since the crunch | Baseline (2025 levels) | ~90% lower, company-reported | Not independently audited | Forbes Australia |
| IPO target window | H2 2026 (per Blackbird) | Widely expected to slip to 2027 | Delayed | Startup Daily, SmartCompany |
Even at the lower US$31 billion internal mark, Canva remains one of the most valuable private software companies to come out of Australia, and it is still profitable on a free cash flow basis by most reporting to date. The story is not that Canva is in trouble the way a company running out of cash is in trouble. It is that the market’s willingness to pay a premium for “AI-native design platform” narrowed the moment investors actually looked at the cost of running that AI at scale, and Google just showed up with a version of the same idea that doesn’t have to answer to Canva’s cap table.
Why the timing reads as more than coincidence
Google did not build Pics in three weeks in response to Canva’s valuation news. Products like this take many months of internal development, and Google had already previewed pieces of the underlying Nano Banana technology earlier in 2026. But the sequencing still matters for how the market reads it. Canva spent mid-August publicly explaining that AI economics forced it to slow down. Three weeks later, the company best positioned to make AI economics look easy, because it owns the model, the cloud infrastructure, and the distribution channel all at once, shipped a rival product at scale.
That is the structural problem Canva’s shareholder letter was really describing, whether or not Perkins framed it that way. A company that licenses or fine-tunes on top of frontier AI models is renting the thing its competitors own outright. Canva’s response, building proprietary in-house models after buying Leonardo.AI, is a reasonable attempt to close that gap. But it takes time, and Google’s Nano Banana lineage has had a running head start inside Google’s own infrastructure, where the marginal cost of an inference call is closer to Google’s internal compute cost than to the price a smaller company pays a frontier lab for API access.
Competitive landscape: it’s not just Canva and Google
Pics is not the only entrant crowding the AI design category this year. Adobe has been pushing Firefly-powered features through Express and the core Creative Cloud suite. Microsoft Designer has been folding into the Microsoft 365 Copilot experience. Canva itself is not standing still: at Canva Create 2026, it introduced Canva AI 2.0, described internally as the biggest product shift since the company launched in 2013, built on what Canva calls its Design Model, aimed at generating fully layered, editable output from a single prompt rather than a flat image.
What makes Google’s entry different from Adobe’s or Microsoft’s is distribution and pricing structure. Adobe and Microsoft both charge directly for their AI design tools as an add-on or a tier upgrade. Google folded Pics into subscriptions many Workspace and Google AI customers already pay for, which removes the “should I pay for this separately” decision that every other AI design tool asks a user to make. That is a meaningfully different go-to-market model, and it is the one most likely to pressure Canva’s free-tier funnel, since a lot of Canva’s growth historically came from free users discovering the product organically before upgrading.
Historical context: Canva’s rise, and the AI cost problem hitting the whole industry
Canva launched in 2013 and built its business on making design accessible to people without design training, using templates rather than a blank canvas. It grew into a company reportedly serving well over 100 million monthly active users and, according to earlier 2026 reporting, crossed roughly US$4 billion in annualized revenue, with growth partly credited to referral traffic from AI chatbots sending users to Canva to complete design tasks the chatbots themselves couldn’t finish. That irony, AI assistants driving traffic to Canva because they couldn’t do the visual work themselves, is part of what makes this year’s reversal notable. The same category of technology that helped Canva grow is now capable of doing more of that work directly, inside tools people already use.
Canva is not alone in discovering that AI inference costs can outrun AI revenue. Across the software industry in 2025 and 2026, multiple companies that rushed to add generative AI features found that the cost of running frontier models at consumer scale ate deeply into margins that used to be near-pure software economics. Canva’s public acknowledgment of the problem, and its decision to slow feature rollout rather than keep burning cash on frontier-model API calls, is one of the more transparent examples of that industry-wide reckoning playing out in a single company’s shareholder letter.
Market impact: what this means for Canva’s IPO clock
The most immediate practical consequence is timing. Blackbird had signaled Canva was “ready” for a US listing in the second half of 2026. A markdown from your own lead investors, followed weeks later by a major new competitor entering general availability, is not the backdrop most companies want heading into roadshow season. Delaying into 2027 buys Canva time to show the AI cost cuts translating into stable or improving margins over a few more quarters, and to show whether Google Pics actually pulls meaningful usage away from Canva’s core template business or ends up serving a different, more casual use case inside Workspace documents.
For employees, the internal valuation used to price equity has already dropped from US$38.9 billion to US$31 billion, which directly affects the value of stock-based compensation package sizes calculated against that mark. For competitors like Adobe and Microsoft, Canva’s public struggle with AI economics is a useful data point: it suggests the AI design category is going to reward companies that either own their model stack outright or can bundle AI costs into a larger, already-profitable subscription base, rather than companies selling AI-design-as-a-standalone-product.
What Canva is doing about it
Canva’s public response has been to lean into cost control rather than feature velocity for the moment. The reported 90% reduction in AI servicing costs, driven by proprietary in-house image and video models plus the integration of Leonardo.AI’s team and technology, is the centerpiece of that response. The company has also continued shipping product outside pure AI generation, including Canva Offline, an expanded Pro design tier, a print-focused Print Shop refresh, and Learn Grid, an education-focused product line, all launched around Canva Create 2026. Spreading growth across areas that are not solely dependent on expensive frontier-model inference is a hedge against exactly the kind of cost shock that triggered the August valuation cut.
Canva also integrated directly with Google’s own Gemini ecosystem earlier in 2026 through a Connected App that links Nano Banana image generation to Canva’s Magic Layers editing tools, according to reporting from DesignRush and TheNextWeb. That is a notable wrinkle: Canva partners with Google’s AI on one front (bringing Nano Banana-generated images into Canva for editing) while now directly competing with a different Google product built on the same underlying model. It illustrates how tangled the AI supply chain has become, where the same lab can be simultaneously a vendor, a platform, and a direct competitor to the same company.
Predictions: what happens next
- Canva’s IPO slips into 2027, but doesn’t disappear. A profitable, multi-billion-dollar-revenue company with a large user base is still an attractive listing candidate; the delay is about optics and pricing power, not viability.
- Google Pics usage will concentrate in Workspace documents rather than pull Canva’s core template users away immediately. The overlay-in-Docs-and-Slides model targets a different moment in the workflow than opening a dedicated design app, so expect gradual erosion of casual use cases first.
- Canva accelerates its own proprietary-model strategy rather than reversing course on AI. Given the 90% cost reduction already reported, expect Canva to keep investing in owning more of its model stack instead of retreating from AI features.
- Adobe and Microsoft respond with tighter AI bundling of their own. Google’s bundled-into-subscription approach with Pics is likely to push Adobe Express and Microsoft Designer toward similar all-in-one pricing rather than separate AI add-on fees.
- Expect more “who owns the model wins” valuation resets across the AI-native software sector. Canva’s markdown is likely to be cited as a template for how investors reassess other AI-feature-dependent startups that don’t control their own underlying model.
What this means for design teams and everyday users right now
For most people, the practical question is simpler than the corporate maneuvering: which tool actually gets a task done faster today. If the work already lives inside a Google Doc or a Slides deck, and the need is a quick generated image or a fast text-in-image translation, Pics removes a context switch that Canva can’t remove, since Canva still requires leaving the document to open a separate app or tab. If the work needs brand consistency, a large template library, print fulfillment, or a team’s existing brand kit, Canva’s depth still outmatches a prompt-first tool that is, as of this month, brand new to general availability and has not yet built out the years of template and workflow tooling Canva has.
Enterprise IT and procurement teams evaluating both tools should note the access structure carefully: Pics is gated behind specific Google AI and Workspace subscription tiers rather than sold as its own line item, so the real comparison for a company already paying for Workspace Business or Enterprise plans may be closer to “free feature we already have” versus “separate Canva subscription,” which changes the calculus considerably compared with a like-for-like price comparison.
Frequently asked questions
What is Google Pics?
Google Pics is an AI-powered image generation and editing tool built on Google’s Nano Banana model. It reached general availability on September 1, 2026, and is available as an overlay inside Google Docs and Slides or as a standalone tool at pics.new.
How much did Canva’s valuation actually drop?
Canva’s lead investors, Blackbird Ventures and AirTree Ventures, cut their valuation estimate by 17%, from roughly US$42 billion to about US$34.9 billion, according to reporting first published by the Australian Financial Review. Canva’s own internal valuation, used for employee equity, separately fell from US$38.9 billion to US$31 billion over the trailing year.
Why did Canva cut its revenue growth forecast?
CEO Melanie Perkins told shareholders the company was lowering its 2026 revenue growth forecast from 30% to 20% because the cost of serving AI tasks through frontier models was too high, and Canva chose to slow new AI feature launches while it worked on reducing those costs.
Is Google Pics free to use?
It is not sold as a separate subscription. Access comes bundled into existing Google AI Pro and Ultra subscriptions, Google AI Pro for Education, and Workspace Business Standard, Business Plus, Enterprise Standard, and Enterprise Plus plans.
Did Google build Pics specifically to compete with Canva?
Reports describe Pics as a direct rival to Canva, Adobe Express, and Microsoft Designer in terms of what it does, but Google has not stated the product was built specifically in response to Canva’s valuation news. The general-availability launch simply landed about three weeks after Canva’s markdown became public.
Has Canva’s AI cost problem actually been fixed?
Canva says it has cut AI servicing costs by around 90% by building proprietary in-house image and video models and integrating technology from its 2025 acquisition of Leonardo.AI. Those figures are company-reported and have not been independently audited.
Will Canva still go public in 2026?
Blackbird Ventures had previously said Canva was “ready” for a second-half 2026 listing, but reporting from Startup Daily and SmartCompany indicates that timeline is now widely expected to slip into 2027 following the valuation markdown.
What model powers Google Pics?
Google Pics runs on Nano Banana, Google’s image generation and editing model, with the newer Nano Banana 2 version supporting precise text rendering, object-level image segmentation, and in-image text translation into 29 languages.
Related Coverage
- How to Use Google Pics: 12 Steps, Nano Banana Pro [2026]
- OpenAI Scientist Warns of AI Risk 3 Days After Astra [2026]
- Grok 4.5 Cuts Coding Costs 80%, Trails Rivals [2026]
- Altman: Next AI Models 'Sobering' After 2-Week Pause [2026]
- OpenAI Halts Training 2 Weeks Over Astra Alarm [2026]
