Akal Cloud

AI agents are provisioning your infrastructure now. AWS made their DevOps Agent generally available two weeks ago. Early customers are reporting 75% lower mean time to resolution on incidents. Impressive. And we are going to see much more of this across every cloud this year. But it raises a question FinOps teams have been asking quietly for months. When an AI agent spins up an r6i.8xlarge at 2 AM to fix a database bottleneck, and that instance is still running two weeks later, who owns the cost? This is not theoretical. We are already seeing teams where automated systems create resources faster than anyone can track the spend. Historically, cost accountability mapped to the team that provisioned the resource. That model starts to break when the team is a Python script in an IAM role. We built Akal Cloud to cut through this ambiguity. Ask Akal "who started that p3.8xlarge last Tuesday?" and it returns the IAM identity, timestamp, and source IP, whether a human or an agent launched it. Across every account in your AWS Organization. No manual CloudTrail spelunking. The tooling can tell you who spent the money. The policy question of who OWNS the cost is one every engineering org needs to answer before the next quarterly review. Vote below. And drop your reasoning in the comments if your org has already figured this out. We are tracking what the community says and will share the results next week. #AkalCloud #AWS #FinOps #AI #DevOps

85% do FinOps. Only 28% are mature. Flexera's 2026 State of the Cloud just confirmed the widest gap the report has ever recorded between adoption and maturity. Most companies have a FinOps Slack channel, a monthly cost review, and a dashboard somebody built last quarter. That's not FinOps. That's cost curiosity. Mature FinOps means three things. You know where every dollar is going, down to the team and the workload. You can trace any anomaly to the exact person who launched the resource. And you act on waste continuously, not once a quarter. Companies in the mature 28% report 40% less cloud waste than everyone else. Across the industry, that gap is worth roughly $270B a year. We built Akal Cloud to close it automatically. Ask Akal "why did our database costs spike last Tuesday?" and it traces the anomaly to the exact instance, timestamp, and IAM user. Across every account in your AWS Organization. No dashboards to build, no CUR exports to wait for. What's your honest answer: 85% bucket, or 28%? #AkalCloud #AWS #FinOps #CloudCosts

A company found $180K/year in a forgotten AWS account last month. It happened on a Thursday. Finance flagged the monthly bill. CTO asked the platform team. Nobody remembered setting up the account. Turns out, an acquired subsidiary's AWS account had been rolling forward for three years. Post-merger, nobody consolidated. No tags. No owner. Just a standing bill. Inside: 12 EC2 instances running at 3% CPU. An RDS cluster nobody had logged into since 2023. Two NAT gateways in empty VPCs. $15K/month. $180K/year. Nothing using any of it. This is not rare. Multi-account AWS orgs grow sideways. Accounts get added for projects, acquisitions, and one-off POCs. Nobody stops paying for them. Spreadsheet rollups of linked accounts do not catch this. We built Akal Cloud because it should not take a Thursday bill escalation to find this. Connect your AWS Organization once. Every linked account, every resource, scanned daily. New accounts auto-discovered. Unified savings view across the entire org. Ask Akal "what accounts haven't had a login in 90 days?" Get the list with monthly spend per account. 14-day free trial on AWS Marketplace. How many linked accounts does your org have? Guess before you look. #AWS #FinOps #MultiAccount #AkalCloud

Your EC2 fleet runs at 14% CPU on average. That's not a pessimistic estimate. AWS Compute Optimizer data across enterprise accounts consistently shows typical EC2 CPU utilization between 10% and 20%. So why is so much capacity sitting idle? Because rightsizing is never urgent, the instances are running. The app is responding. Nobody gets paged. Meanwhile, a fleet of m5.2xlarge machines that could be m5.large is quietly billing three sizes larger than needed for months. The math gets uncomfortable fast. 50 instances averaging one size too large on a $60K/mo EC2 bill is roughly $15K/mo in pure capacity waste. Every month. Until someone looks. We built Akal Cloud to make "someone" unnecessary. Ask Akal "which instances are under 5% CPU across all accounts?" and you get each one with the current cost and the recommended size. Ask "which m5 instances haven't hit 20% CPU in the last 30 days?" and you get a cleanup list. Across an entire AWS Organization. Not one account at a time. Most teams save more in the first week than they pay us in a year. When was the last time your team ran a rightsizing pass across every account in your org? #AkalCloud #AWS #FinOps #EC2 #CloudCosts

5% of your AWS bill is buying nothing. A recent analysis of enterprise AWS accounts found that 5.25% of the average bill goes to unattached EBS volumes and orphaned snapshots. Across the industry that adds up to $2.6B a year in pure waste. The pattern is always the same. A team spins up a test environment, detaches the volume, forgets the snapshot policy is still running. A year later the volume is still there. The snapshots have quietly multiplied into thousands. On a $200K/mo AWS bill, that's $10K every month going to storage attached to nothing. It never shows up in a cost review because storage feels too small to investigate. We built Akal Cloud to find exactly this. Ask Akal "which EBS volumes are unattached and how much are they costing?" and it returns each volume, the size, the region, and the monthly cost. Same for snapshots older than 90 days with no parent. Across every account in your AWS Organization, not one at a time. Two minutes of cleanup usually pays for a year of the tool. What's the largest "nobody touched this in a year" resource your team has found? #AkalCloud #AWS #FinOps #EBS #CloudWaste

A team spun up eight p3.8xlarge instances for a machine learning training job. The job finished in three days. The instances kept running for six weeks. Nobody noticed because GPU costs were buried inside a $2M monthly AWS bill. The training team assumed infrastructure would handle it. The infrastructure team assumed the ML engineers would clean up after themselves. Total cost of those six idle weeks: roughly $140,000. This isn't unusual. We see this pattern constantly. The gap between "who creates resources" and "who pays for resources" is where cloud waste lives. And it's getting worse. GPU instances cost 5-10x standard compute. A single forgotten ml.p4d.24xlarge runs over $22,000 per month. When the most expensive resources are also the easiest to forget about, the math gets ugly fast. The fix isn't better budgets. It's real-time visibility into what's running and what it costs, tied to who started it. At Akal Cloud, Akal answers that question directly. "Who launched that expensive instance?" returns the IAM user, timestamp, and source IP. "What resources are running that nobody has touched in 30 days?" lists each one with the monthly cost. The $140,000 in idle GPUs would have been flagged on day one. How does your team handle cleanup after ML training jobs? #AkalCloud #AWS #FinOps #MachineLearning #CloudCosts

Cloud waste just went up for the first time in five years. Flexera's 2026 State of the Cloud report puts it at 29%. That breaks a long stretch where waste hovered around 27%. The culprit? AI workloads. Teams are spinning up GPU instances for training and inference, and nobody has figured out how to govern the spend yet. Average GPU utilization across enterprise workloads is just 23%. That means 77% of provisioned GPU capacity is sitting idle at any given time. Meanwhile, 98% of FinOps teams now say they manage AI spend. Two years ago it was 31%. The scope grew faster than the tooling. The irony is that most of this waste is visible. Idle instances, over-provisioned clusters, unused commitments. It shows up in the data every single day. The problem isn't detection. It's that nobody is looking at it in real time. We built Akal Cloud for exactly this. Ask Akal "what resources are running that nobody has touched in 30 days?" and it returns each one with the monthly cost attached. 150+ checks across 32 AWS services, running daily against your live account. Cloud waste going up should worry every engineering leader. But it also means the savings opportunity just got bigger. What percentage of your cloud spend do you think is actually wasted? #AkalCloud #AWS #FinOps #CloudCosts #CloudWaste

AWS just made their DevOps Agent generally available. It investigates incidents, resolves issues, and manages infrastructure. Early customers report 75% lower mean time to resolution. That's impressive. But it raises a question we keep hearing from FinOps teams. When AI agents can spin up and modify infrastructure autonomously, who is responsible for the cost of what they create? Right now, 78% of FinOps practices report to the CTO or CIO. Cost accountability usually maps to the team that provisioned the resources. But if an AI agent provisions them during an incident at 2 AM, which team's budget does that hit? This isn't theoretical. We're already seeing organizations where automated systems create resources faster than anyone can track the spend. At Akal Cloud, we built Akal to answer exactly this kind of question. "Who started that p3.8xlarge last Tuesday?" returns the IAM user, timestamp, and source IP, whether it was a person or an automated process. Who should own cloud costs in your org? A) Engineering, they create the resources B) Finance, they manage the budget C) A dedicated FinOps team D) The platform team #AkalCloud #AWS #FinOps #DevOps #CloudCosts

Most teams buy Savings Plans and assume they're covered. They're not. We looked at dozens of AWS accounts this quarter. The pattern is consistent: teams commit to Compute Savings Plans for EC2 and Lambda, then stop thinking about it. Meanwhile, RDS, OpenSearch, Redshift, ElastiCache, and Neptune run on-demand pricing, month after month, with no commitment discount applied. AWS just expanded Database Savings Plans to cover OpenSearch and Neptune. That's good news, but it also means the coverage gap just got wider for teams that aren't paying attention. Here's what we keep seeing: Compute Savings Plans covering 60-70% of eligible EC2 spend. Database workloads running at full on-demand rates. No one auditing whether the commitment level still matches actual usage after workload migrations. The savings on the table are real. Reserved Instances and Savings Plans can cut 40-72% off stable workloads. But only if someone is actually tracking which workloads are covered and which aren't. We built Akal Cloud to catch exactly this. Ask Akal "what would we save with Reserved Instances on our RDS fleet?" and it analyzes your usage patterns against current pricing, from live data. When was the last time someone audited your Savings Plan coverage? #AkalCloud #AWS #SavingsPlans #FinOps #CloudCosts

AWS just raised EC2 Capacity Block pricing by 15%. Every region. Every ML instance type. p4d, p5, trn1, all of them. If your team reserves GPU capacity for training or inference, your next invoice is going up. No grandfathering. This matters because most teams budget GPU costs based on last quarter's rates. A 15% jump mid-cycle breaks forecast accuracy and blows through committed budgets before anyone notices. Three things to check right now: 1. Review your Capacity Block reservations. Are you actually using the full reserved window, or paying for GPU hours that sit idle overnight and on weekends? 2. Compare on-demand vs Capacity Block vs Savings Plans for your ML workloads. The math just changed. For bursty training jobs, on-demand might now be cheaper than reserving. 3. Check your inference fleet. If you scaled up GPU instances for a model launch and never scaled back down, that 15% compounds on waste you already had. We built Akal Cloud to surface exactly this kind of cost drift. Ask Akal "what are we spending on GPU instances this quarter vs last?" and it shows the delta from live data, not a CUR file from three days ago. When was the last time you re-evaluated your GPU reservation strategy? #AkalCloud #AWS #EC2 #MachineLearning #FinOps