This is a matter of how reliable do you want it with Arrhenius Law and using Ohm's Law for thermal resistance of the hottest transformer and your enclosure effective thermal resistance. The linear air hot-surface speed matters far more than the internal ambient and CFM volume air flow. Most PC towers are poorly designed for this.
You must choose max hot spot of Ferrite as the most critical element to prevent thermal runaway. Typically I found from doing this design before in a high volume product. The OEM suppliers design for open air, not enclosed, which they have no control over. They measure internal ambient maybe, but not the hotspot. So I decided to do this and found that 2 1U (38mm?) high 1W fans easily remove fans regulated with thermistor and Amplifier insulated on chassis in air path. My chassis was rack mount, sealed lid ends vented with a custom Mylar spoiler on intake to make turbulent flow then laminar flow in exit. That made the hotspot 20'C cooler than without the spoiler. The fans only needed a little power above about 50% at full load at room temp so it could survive 40'C ambient. YMWV
Up to 90% efficiency means best case of 150 W. which could be some value below max power and might mean 85% at 150W (est.) which means 23W if heat on the ferrite XFMR and FETS shared unequally. I suggest you measure the hotspot and ambient heat rise in a box with a fan and see the temp rise above room on the hottest part. I chose 20'C rise max. Above 40'C internal ambient, initially before I figured out this stuff, it was a 45'C rise.
- then you have to add the load heat of internal dissipation.
You must also consider fan noise and safety requirements. UL mandated some coke spill requirement and crush test for the lid. also it was tested and passed all conducted emissions and UL with caveats approved for painted welds on ground studs.
More details
BTW I epoxied a 10k thermistor to the XFMR and tested HIPOT with DC earth bonded and found 6 failure exceptions in their production as they test HIPOT easier with DC floating. The XFMR capacitance causes more HIPOT primary stress and failures when secondary is earth bonded .