Depends on whom is calculating the costs.
Insurance companies profit every time you need healthcare, and can find a plausible reason to ration it or deny it completely. Most insurance companies now do pre-authorization for services based on the diagnosis codes. By making diagnosis coding more granular, they have more of an opportunity to save money prior to dispensing care, or by denying your claim after the fact, by saying "You had diagnosis X, which our contract with your provider does not cover for procedure Y". Conservatives refer to this process as a death panel. Liberals refer to it as "bending the cost curve", which has a more pleasant ring to it, even if the net effect is that you don't get treatment in a timely manner, if at all. Please select your preferred side of the political spectrum and rage accordingly.
Providers' costs will probably be a wash. Most providers already have someone who is a Certifier Professional Coder on staff, or they contract it out to a third party coding company. Strangely, these people refer to themselves as Coders, which tends to confuse the hell out of the IT staff. This individual is tasked with reading the pre- and post-procedure diagnosis and procedure report, and assigning a set of CPT and ICD codes to the encounter. These are, in theory, used for clinical purposes, but the main reason procedure and diagnosis codes exist is that they form the terms of the contract between payer and provider. This coding process tends to have the same precision as a group of witches divining the future from cat entrails, because of a wonderful little concept called "unbundling". Let's say that an anesthesiologist is performing a spinal block in which they inject narcotics and steroids into multiple levels in your back. Some combinations of spinal blocks are issued their own procedure codes, whereas other spinal blocks can be billed for separately. If a coder fails to group the separate procedure codes into a proper bundle, or fails to justify the code with the correct diagnosis, CMS and an entire army of subpoena-wielding acronyms will crawl up their butt screaming "FRAUD!!!", because frequently the total of several individual codes will pay much more than a single bundled code. With a more granular coding system, the coder may need to spend a bit more time per report finding the right level of detail, but in theory the number of unpaid visits and claim denials will go down, because they and the insurance company now have a more accurate agreement on what diagnosis will justify the treatment they are providing. ICD-10 covers the diagnosis portion of the coding system, but rest assured, the AMA will soon be confusing the hell out of us with an equally arcane set of new procedure codes.
Incidentally, the procedure codes can and do create some perverse incentives in healthcare. For example, if an orthopedic surgeon is performing a knee procedure with an MCL repair and doing it laproscopically, she is actually paid MORE if she makes two incisions than if she makes one. Reason? There is a single code for "Knee arthroscopy with MCL repair", and separate codes for arthroscopy and MCL repair as their own procedure. Medically, it's safer if you do both procedures with a single incision because of lowered infection risk and less bleeding, but the the difference between the bundled code and the unbundled code is that second incision. So the coding system actually creates a financial incentive to perform a second incision. This is especially true if the procedure is done in an outpatient facility that the surgeon happens to have ownership in.
Universities, research companies, EMR vendors, and biotech firms using big data analysis to perform studies will consider this granularity a godsend, because they won't have to squint as hard at the data to tease things out. Whether you, the patient, want your record to be teased without your knowledge or consent is a debate for the future, but the medical benefits of being able to data-mine everyone's record will likely far outweigh the privacy concerns. Without a centralized way of accessing medical records in an anonymized manner, this will not happen anytime soon.
EMR firms will also be big winners. As of today, electronic medical records are not directly processed by the government, because there is no federal standard for medical record interchange. Some states, notably Massachusetts with their HiWAY program, have either built or are planning to build a statewide medical record exchange, so that each EMR vendor will only have to deal with one format in and out, and the exchange will handle the details of transferring records between providers (e.g. you have a lab test done that your surgeon needs to look at, but both are separate business entities). Although it isn't on the legislative roadmap yet, sooner or later Congress critters will get it through their Kevlar-plated skulls that this might be a good thing for the feds to subsidize and/or standardize, so that EMR vendors have one standard to deal with, rather than 50. Having a reasonably granular set of codes means that describing an encounter will be easier to do in a standardized manner, which in turn should lead to more portability. Of course, with every seismic shift in healthcare IT comes a slightly delayed tsunami of billable hours from IT consultants "helping" you navigate the transition, so providers will lose as much as EMR vendors will win, if not more in terms of productivity losses associated with reimplementations.
Patients, as always, will lose money, because the cost difference between what the provider pays and what the insurance company thinks they will pay will likely increase.